JP5004736B2 - Image processing apparatus and image processing program - Google Patents

Description

  The present invention relates to an image processing apparatus and an image processing program for processing time-series images.

  In recent years, in the field of endoscopes, an imaging function that captures an image inside a subject, a transmission function that wirelessly transmits image information captured by an imaging unit, and the like are housed in a capsule-shaped case. A swallowable capsule endoscope has been proposed. This capsule endoscope is swallowed from the patient's mouth, which is the subject for examination, introduced into the subject, and naturally discharged, such as the esophagus, stomach, small intestine, large intestine, etc. It moves inside the organ according to its peristaltic movement. Then, while moving inside the body, images in the body are sequentially captured at, for example, 2 to 4 (frame / sec), and the captured image information is wirelessly transmitted to a receiving device outside the body. Images inside the subject taken by the capsule endoscope and received by the external receiving device are sequentially displayed in time series on a diagnosis workstation or the like and confirmed by an observer such as a doctor.

  This capsule endoscope captures an enormous number of images. For this reason, in a diagnosis workstation or the like, a process for detecting an image suspected of the presence of an abnormal part such as bleeding as an image to be observed is performed from among the captured images. This burden is reduced (for example, refer to Patent Document 1). In this Patent Document 1, for each unit section obtained by dividing an image into regions, a feature amount related to a color tone is calculated using pixel information, and a process for determining whether or not an abnormal portion is included is performed, whereby the presence of an abnormal portion is detected. Suspicious images are detected.

JP 2005-192880 A

  By the way, in the technique of Patent Document 1, when determining whether or not an abnormal part is included, the abnormal part is determined for all unit sections. However, if an abnormal part is detected even at one place in the image, this image is an image that should be confirmed by a doctor or the like. Therefore, in order to determine whether the image includes the abnormal part, the abnormal part is always detected for all unit sections. In some cases, the processing time may be increased.

  In addition, the detection of the abnormal portion in the image is performed in a predetermined order, for example, by performing raster scanning in order from the upper left of the image. For this reason, in conventional abnormal part detection, for example, even when an image including an abnormal part in the lower right in the image is continuous in time series, the abnormal part is always detected by raster scanning sequentially from the upper left of the image. As a result, the processing time may be increased.

  The present invention has been made in view of the above, and suppresses processing to be performed when detecting an image including an abnormal part from time-series images to the minimum necessary, and detects an image including the abnormal part in a short time. An object of the present invention is to provide an image processing apparatus and an image processing program that can be used.

  In order to solve the above-described problems and achieve the object, an image processing apparatus according to the present invention is an image processing apparatus that processes a time-series image composed of a plurality of images photographed in time series, In the images constituting the time-series image, priority order setting means for setting the priority order for the pixel or area to be processed, and the abnormal part of the subject in order from the pixel or area having the higher priority order set by the priority order setting means. An abnormal part detecting means for sequentially detecting a corresponding region from the image for a plurality of images constituting the time-series image; and the abnormal part detecting means based on a result of the detecting process And a process stop control means for controlling the stop of the detection process in each image according to the above and a detection result recording means for recording the result of the detection process.

  Further, an image processing program according to the present invention allows a computer that processes a time-series image composed of a plurality of images taken in time series to process a pixel or region to be processed in an image constituting the time-series image. A priority order setting step for setting a priority order for, and a detection process for detecting, from the image, an area corresponding to an abnormal part of the subject in order from a pixel or area having a higher priority order set by the priority order setting step, An abnormal part detection step sequentially performed for a plurality of images constituting the time series image, and a process stop for controlling stop of the detection process in each image by the abnormal part detection step based on a result of the detection process A control step and a detection result recording step for recording the result of the detection process are executed.

  According to the image processing apparatus of the present invention, it is possible to perform processing for detecting abnormal portions in order from pixels or regions having higher priorities, and stop the processing for detecting abnormal portions for the processing target image according to the result. Therefore, it is possible to suppress the processing related to the detection of the abnormal part to the minimum necessary, and to detect an image including the abnormal part from the time series images in a short time.

  DESCRIPTION OF EMBODIMENTS Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the drawings. In the embodiment described below, an image processing device that processes a plurality of in-vivo time-series images captured in time series by a capsule endoscope will be described. The processable time-series image is not limited to the time-series image obtained by imaging the living body.

(Embodiment 1)
First, the first embodiment will be described. FIG. 1 is a schematic diagram illustrating an overall configuration of an image processing system including the image processing apparatus according to the first embodiment. As shown in FIG. 1, the image processing system is wirelessly transmitted from a capsule endoscope 10 that captures an image inside a subject 1 (hereinafter referred to as an “in-vivo image”) and the capsule endoscope 10. A receiving device 30 that receives an in-vivo image, an image processing device 70 that processes and displays the in-vivo image captured by the capsule endoscope 10 based on the in-vivo image received by the receiving device 30, and the like Consists of. For example, a portable recording medium (portable recording medium) 50 is used for transferring image data between the receiving device 30 and the image processing device 70.

  The capsule endoscope 10 has an imaging function, a wireless function, and the like. The capsule endoscope 10 is swallowed from the mouth of the subject 1 and introduced into the subject 1, and sequentially in vivo images while moving in the body cavity. Take an image. Then, the captured in-vivo image is wirelessly transmitted outside the body.

  The receiving device 30 includes a plurality of receiving antennas A1 to An, and receives in-vivo images that are wirelessly transmitted from the capsule endoscope 10 via the receiving antennas A1 to An. The receiving device 30 is configured so that the portable recording medium 50 can be freely attached and detached, and sequentially stores the received image data of the in-vivo image in the portable recording medium 50. In this way, the in-vivo image inside the subject 1 captured by the capsule endoscope 10 is accumulated in the portable recording medium 50 in time series by the receiving device 30 and stored as a time series image.

  Here, the imaging rate of the in-vivo image by the capsule endoscope 10 is normally 2 to 4 (frame / sec), and a huge number of in-vivo images of tens of thousands or more during the stay time in the body of about 8 hours. An image is taken. For this reason, the image data of the in-vivo image captured by the capsule endoscope 10 is compressed by a compression technique such as JPEG using the DCT encoding method and stored in the portable recording medium 50. In the compression processing using this DCT encoding method, discrete cosine transform (DCT) is performed for each predetermined unit partition (usually 8 × 8 pixel block) to convert image data into frequency components, and DCT which is a conversion coefficient thereof. Image data is compressed by quantizing the coefficient for each unit section. In JPEG, after an imaging signal is converted into a luminance signal (Y signal) and a color difference (u signal, v signal), DCT processing is performed on each signal.

  The receiving antennas A1 to An are configured by, for example, loop antennas, and are distributed at predetermined positions on the body surface of the subject 1 as shown in FIG. Specifically, for example, they are distributed and arranged at positions on the body surface corresponding to the passage route of the capsule endoscope 10 in the subject 1. Note that the receiving antennas A1 to An may be arranged in a distributed manner on a jacket worn by the subject 1. In this case, the receiving antennas A1 to An are predetermined on the body surface of the subject 1 corresponding to the passage route of the capsule endoscope 10 in the subject 1 when the subject 1 wears this jacket. Placed in position. Moreover, the receiving antenna should just be arrange | positioned 1 or more with respect to the subject 1, and the number is not limited.

  The image processing apparatus 70 is for a doctor or the like to observe and diagnose an in-vivo image captured by the capsule endoscope 10, and is realized by a general-purpose computer such as a workstation or a personal computer. The image processing apparatus 70 is configured so that the portable recording medium 50 can be freely attached and detached, processes in-vivo images constituting a time-series image stored in the portable recording medium 50, and displays such as an LCD or an ELD. Are displayed sequentially in chronological order. In-vivo images of body cavities show mucous membranes in the body cavities, contents floating in the body cavities, bubbles, etc., and sometimes noticeable abnormal parts such as bleeding sites, lesion sites, abnormal sites, etc. . The image processing device 70 detects the position in the in-vivo image where the abnormal part to be noted is reflected, and performs a process of displaying the position separately from the other.

  FIG. 2 is a block diagram illustrating a functional configuration of the image processing apparatus 70 according to the first embodiment. In Embodiment 1, the image processing apparatus 70 includes an image acquisition unit 710, an input unit 720, a display unit 730, a storage unit 740, an image processing unit 750, and a control unit 770 that controls each unit of the apparatus. .

  The image acquisition unit 710 acquires an in-vivo image that forms a time-series image captured by the capsule endoscope 10 and stored in the portable recording medium 50 by the receiving device 30. The medium 50 is detachably mounted, and the image data of the in-vivo image stored in the mounted portable recording medium 50 is read and acquired. The image acquisition unit 710 is realized by, for example, a read / write device corresponding to the type of the portable recording medium 50. The in-vivo image acquired by the image acquisition unit 710 is output to the image processing unit 750 via the control unit 770, and is decompressed and restored by a procedure reverse to the compression processing. The acquisition of time-series in-vivo images captured by the capsule endoscope 10 is not limited to the configuration using the portable recording medium 50. For example, a hard disk is used instead of the image acquisition unit 710. The time series in-vivo image imaged by the capsule endoscope 10 may be stored in advance in the hard disk. Alternatively, a server may be separately installed instead of the portable recording medium 50, and time-series in-vivo images may be stored in advance on this server. In this case, the image acquisition unit is configured by a communication device or the like for connecting to the server, and connected to the server via the image acquisition unit to acquire time-series in vivo images from the server.

  The input unit 720 is realized by, for example, a keyboard, mouse, touch panel, various switches, and the like, and outputs an operation signal corresponding to the operation input to the control unit 770. The display unit 730 is realized by a display device such as an LCD or an ELD, and displays various screens including a time-series in-vivo image display screen under the control of the control unit 770.

  The image processing unit 750 is realized by hardware such as a CPU. The image processing unit 750 extends the image data of the in-vivo image read and acquired from the portable recording medium 50 by the image acquisition unit 710, and the restored in-vivo image is displayed. Various arithmetic processes for displaying on the display unit 730 are performed. The image processing unit 750 includes a priority setting unit 753 corresponding to a priority setting unit, an abnormal part detection unit 755 corresponding to an abnormal part detection unit, and an abnormal part detection process stop control unit 757 corresponding to a process stop control unit. Including.

  The priority order setting unit 753 applies to each unit section, which is a processing unit of the DCT process, based on the result of the abnormal part detection process performed using the previous (for example, immediately preceding) in-vivo image in the time series as the processing target image. Priority order setting processing for setting a priority order when the abnormal portion detection unit 755 performs the abnormal portion detection processing is performed.

  The abnormal part detection unit 755 sequentially sets each unit section as a processing target (hereinafter referred to as “processing unit section” as appropriate) according to the priority set by the priority setting unit 753, and causes a bleeding site, a lesion site, an abnormal site, and the like. An abnormal part detection process (detection process) for detecting a unit block suspected of being an abnormal part as an abnormal part is performed. The abnormal part detecting unit 755 obtains the feature amount of the processing unit section from the DCT coefficient obtained for each unit section at the time of the expansion process of the processing target image. Then, the abnormal part detection unit 755 performs abnormal part detection based on the obtained feature amount of the processing unit section, and calculates a certainty factor that is an index of detection reliability.

  Here, the DCT coefficient will be described. FIG. 3 is a diagram illustrating DCT coefficients acquired for each unit partition during the decompression process. For a unit block of 8 × 8 pixel blocks, 64 (= 8 × 8) DCT coefficients as shown in FIG. 3 are obtained for each of the luminance signal Y, the color difference signal u, and the color difference signal v. (Total 192). Further, the 64 DCT coefficients are 0th-order frequency components, and each of the DC components corresponding to the average value information in the corresponding unit section (“DC” in FIG. 3) is different in direction and frequency in the unit section. It is composed of 63 AC components (a plurality of “ACs” in FIG. 3) corresponding to information obtained when the frequency components are decomposed. For this reason, by referring to the DC component of each signal of Y, u, v, it is possible to determine the difference in luminance and color of the corresponding unit section, and to respond by referring to the AC component of each signal. Differences in texture (pattern) of unit sections can be discriminated. Of course, instead of referring to the DCT coefficient itself, a numerical value calculated secondarily based on the DCT coefficient may be used.

  In an in-vivo image, in general, an area showing a bleeding site where blood is flowing out differs greatly in color from surrounding mucosal areas depending on the blood. In addition, a region that reflects an abnormal site that is peeled off with villi and white is greatly different in color and mucous membrane pattern from a normal mucous membrane region. The content area and the bubble area are also greatly different from the normal mucous membrane area in color and surface pattern. For this reason, by using DCT coefficients including color and surface pattern information as feature quantities, abnormal areas with bleeding, normal mucous membrane areas, contents areas with stool etc. floating, and bubbles are generated. It is possible to distinguish a bubble area or the like. Therefore, in the first embodiment, using such characteristics, the DCT coefficient obtained at the time of the expansion process is used as the feature amount. Then, the abnormal part is detected for each unit section based on the feature amount.

  Specifically, the feature amount is vector data having each DCT coefficient in the unit partition as an individual element. The abnormal part detection unit 755 calculates, for each unit section, a feature amount having the luminance signal (Y signal) and the DCT coefficients of the two color difference signals (u signal and v signal) as elements. The calculated feature amount is 64 × 3 = 192 dimensional vector data. Note that this feature amount does not need to have all the DCT coefficients in the unit section as elements, and any DCT coefficient may be selected for each signal from within the unit section to be processed as an element of the feature amount. In addition, the feature quantity is generated using the DCT coefficient itself as an element, but instead of using the DCT coefficient itself as an element, a constant multiple of DCT coefficient, a constant number, a square value of DCT coefficient, or a plurality of DCT coefficients. An arithmetic value that is secondarily calculated based on the DCT coefficient, such as an average value of the coefficient, can be used as an element.

  The abnormal part detection process stop control unit 757 performs control to stop the abnormal part detection process by the abnormal part detection unit 755 based on the certainty factor calculated by the abnormal part detection unit 755.

  The control unit 770 is realized by hardware such as a CPU. The control unit 770 performs image processing based on image data of time-series in vivo images acquired by the image acquisition unit 710, operation signals input from the input unit 720, programs and data stored in the storage unit 740, and the like. Instructions to each unit constituting the apparatus 70, data transfer, and the like are performed, and the overall operation of the image processing apparatus 70 is comprehensively controlled. In addition, the control unit 770 displays the in-vivo image in which the abnormal part is detected as a result of the abnormal part detection process by the abnormal part detection unit 755 among the in-vivo images constituting the time-series image stored in the portable recording medium 50. In addition, an image display processing unit 771 that performs processing of sequentially displaying on the display unit 730 is included. The image display processing unit 771 is a functional unit corresponding to display processing means.

  The storage unit 740 is realized by various IC memories such as ROM and RAM such as flash memory that can be updated and stored, a built-in hard disk connected by a data communication terminal, an information recording medium such as a CD-ROM, and a reading device thereof. A program relating to the operation of the image processing apparatus 70, a program for realizing various functions provided in the image processing apparatus 70, data relating to the execution of these programs, and the like are stored. Further, the storage unit 740 causes the image processing unit to function as the priority order setting unit 753, the abnormal part detection unit 755, and the abnormal part detection process stop control unit 757, and causes the control unit 770 to function as the image display processing unit 771. Image processing program 741, priority order setting data 743, and abnormal part detection data 745 are stored.

  The priority order setting data 743 indicates the result of the priority order setting process by the priority order setting unit 753, and is stored in the storage unit 740 as a priority order recording unit. For example, the priority order setting unit 753 allocates serial numbers in order from the unit partition having the highest priority according to the priority order set for each unit partition, and generates priority order setting data 743 by associating the allocated serial number with the identification number of the unit partition. Then, by storing in the storage unit 740, the result of the priority order setting process is recorded.

  The abnormal part detection data 745 indicates the result of the abnormal part detection processing performed on each in-vivo image by the abnormal part detection unit 755, and is stored in the storage unit 740 serving as a detection result recording unit. FIG. 4 is a diagram illustrating a data configuration example of the abnormal part detection data 745. As shown in FIG. 4, the abnormal part detection data 745 is a data table in which an abnormal part detection result and an abnormal part detection flag are set in association with an image number. Is set. The image number indicates the time-series order of each in-vivo image constituting the time-series image stored in the portable recording medium 50. For example, the in-vivo image received by the receiving device 30 is stored in the portable recording medium 50. When the image acquisition unit 710 acquires an in-vivo image from the portable recording medium 50 in the image processing apparatus 70, the image number is assigned to each in-vivo image according to the acquisition order. May be assigned. In the processing unit section, an identification number of the unit section subjected to the abnormal part detection process is set. As the certainty factor, the certainty factor obtained in the abnormal part detection processing performed for the corresponding processing unit section is set. The abnormal part detection flag is flag information indicating whether or not the corresponding processing unit section is detected as an abnormal part, and “OFF” is set as an initial value. As a result of the abnormal part detection process, when the processing unit section is detected as an abnormal part, the abnormal part detection flag is changed to “ON”. The abnormal part detection data 745 may be written in the portable recording medium 50.

  Next, the flow of processing performed by the image processing apparatus according to the first embodiment will be described with reference to FIGS. FIG. 5 is an overall flowchart illustrating a processing procedure performed by the image processing apparatus 70 according to the first embodiment. Note that the processing described here is realized by the operation of each unit of the image processing apparatus 70 in accordance with the image processing program 741 stored in the storage unit 740.

  First, the image processing unit 750 initializes the image number i of the in-vivo image to be processed (processing target image) to “i_start” (step a1). i_start is the image number of the first in-vivo image in the time-series image. Then, the image processing unit 750 acquires the image data of the in-vivo image I (i) with the image number i via the image acquisition unit 710 and the control unit 770 (step a3). At this time, the image processing unit 750 decompresses the acquired image data and restores the in-vivo image I (i). The DCT coefficient obtained for each 8 × 8 pixel block, which is a unit block in JPEG, at the time of this expansion processing is temporarily held in the storage unit 740. At this time, the image may not be completely restored, and the abnormal part detection process may be performed by simply reading the DCT coefficient, and when the image is confirmed later, only the image having the abnormal part may be restored and displayed. As a result, it is not necessary to perform the inverse DCT process on an image having no abnormal part, and the processing time can be increased.

  Next, the priority order setting unit 753 performs priority order setting processing (step a9). Subsequently, the abnormal part detecting unit 755 initializes the priority order k of the unit section to be processed to “1” (step a11), and sets the unit section having the priority order k in the processing target image I (i) as a processing unit. An abnormal part detection process is performed as a section (step a13). Then, the abnormal part detection process stop control unit 757 determines whether or not to stop the abnormal part detection process by the abnormal part detection unit 755 based on the result of the abnormal part detection process. If the abnormal part detection process stop control unit 757 determines that the abnormal part detection process is to be stopped (step a15: Yes), the abnormal part detection unit 755 detects the abnormal part corresponding to the processing unit section of the priority order k. The flag is set to “ON”, the abnormal part detection data 745 is updated, and this processing unit section is recorded as the abnormal part detection position (step a17). Thereafter, the process proceeds to determination of the presence / absence of the in-vivo image I (i) to be processed next (step a23). On the other hand, when the abnormal part detection process stop control unit 757 determines that the abnormal part detection process is not stopped (step a15: No), the abnormal part detection unit 755 has performed the abnormal part detection process for all unit sections. Determine whether or not. And when it has performed (step a19: Yes), it transfers to step a23. When there is an unprocessed unit section (step a19: No), the abnormal part detection unit 755 increments the priority order k to k = k + 1 (step a21), and performs the processes of steps a13 to a19.

  In step a23, the image processing unit 750 determines whether or not there is an in-vivo image I (i) to be processed next by i = i_end. If i ≠ i_end (step a23: No), the image processing unit 750 increments the image number i to i = i + 1 (step a25), and returns to step a3. On the other hand, if i = i_end (step a23: Yes), the process proceeds to step a27. In step a27, the image display processing unit 771 executes image display processing, and the in-vivo images belonging to the image numbers i_start to i_end, in which the abnormal part is detected as a result of the abnormal part detection process in step a13, are displayed. A process of sequentially displaying on the display unit 730 as a display target is performed.

  Next, the priority setting process in step a9 in FIG. 5 will be described. FIG. 6 is a flowchart showing a detailed processing procedure of the priority order setting process. In this priority order setting process, the priority order setting unit 753 first refers to the abnormal part detection data 745 and displays an abnormal part for the in-vivo image I (i-1) immediately before the processing target image I (i) in time series. The result of the detection process is acquired (step b1). When “ON” is set in the abnormal part detection flag corresponding to the acquired image number i−1, the priority setting unit 753 indicates that the abnormal part is in the in-vivo image I (i−1). It determines with having detected (step b3: Yes), and sets a priority to each unit division according to the result of this abnormal part detection process (step b5). At this time, the priority order setting data 743 is updated based on the set priority order. Since the capsule endoscope 10 repeats the imaging operation at an imaging rate of 2 to 4 (frame / sec), when an abnormal part appears in the captured in-vivo image, the in-vivo image that is close in time series is displayed. However, there are many cases where this abnormal part is reflected, and there is a high possibility that this abnormal part is reflected in the same position. Therefore, the priority order setting unit 753 sets the priority order of each unit section based on the abnormal part detection position detected in the immediately preceding in-vivo image in time series. FIG. 7 is an explanatory diagram for explaining a priority setting method, and FIG. 7-1 shows an abnormal part detection position detected in the immediately preceding in-vivo image I (i-1) in time series. FIG. 7-2 shows the priority order set for each unit section in the processing target image I (i) based on the abnormal part detection position of FIG. As shown in FIG. 7A, when the abnormality detection flag is “ON” and the processing unit section recorded as the abnormal part detection position is the unit section E10, as shown in FIG. The priority order of the unit section E10 in the image I (i) is set to “1” which is the highest, and the priority order of other unit sections is set so that the unit section closer to the unit section E10 is higher.

  On the other hand, as illustrated in FIG. 6, the priority order setting unit 753 determines that an abnormal part is not detected if “ON” is not set in the abnormal part detection flag corresponding to the image number i−1 ( Step b3: No). In this case, the priority order setting unit 753 reads the priority order setting data 743, and the priority order of each unit section held in the priority order setting data 743, that is, the in-vivo image I which is the previous processing target image. In accordance with the priority order set for (i-1), a priority order is set for each unit partition (step b7). For example, when an abnormal part appears in one in-vivo image, this abnormal part is not reflected in the next in-vivo image in time series because the posture of the capsule endoscope 10 has changed. Furthermore, it may be reflected in the next in-vivo image. Therefore, even when an abnormal part is not detected in the previous in-vivo image in time series, the abnormal part detection position detected in the previous in-vivo image in time series, that is, the processing target in the previous time Abnormal part detection position detected in the in-vivo image processed as an image (in the in-vivo image processed as the processing target image before the abnormal part in the previous in-vivo image) Or an abnormal part may be detected in the periphery. For this reason, in the first embodiment, the priority set for the immediately preceding in-vivo image I (i-1) in time series may be used, and the unit section of the abnormal part detection position detected last time may correspond to the abnormal part. Is specified as a unit section with a priority, and a priority order is set. If the priority order has been set, the process returns to step a9 in FIG. 5 and then proceeds to step a11.

  Next, the abnormal part detection process by step a13 of FIG. 5 is demonstrated. FIG. 8 is a flowchart showing a detailed processing procedure of the abnormal part detection processing. In this abnormal part detection process, the abnormal part detection unit 755 acquires the DCT coefficient of the processing unit section temporarily held in the storage unit 740 (step c1). Subsequently, based on the DCT coefficient in the processing unit section acquired in step c1, the abnormal part detection unit 755 calculates the feature amount of the processing unit section (step c3). Then, based on the acquired feature amount, the abnormal part detecting unit 755 calculates a certainty factor in detecting the abnormal part for the processing unit section (step c5). Then, the process returns to step a13 in FIG. 5, and then proceeds to step a15.

  Here, a method for detecting an abnormal part will be described. In the first embodiment, for example, the DCT coefficient of the abnormal part is acquired as a sample for each type in advance, and the feature amount distribution information of the abnormal part calculated based on the acquired DCT coefficient is prepared as teacher data. . FIG. 9 is a diagram illustrating an example of the teacher data of the feature amount of the abnormal part. The teacher data is prepared for each type of abnormal part such as a bleeding part, a lesion part, and an abnormal part. FIG. 9 illustrates three abnormal part types of abnormal types A, B, and C. In FIG. 9, the teacher data is illustrated in a plan view, but as described above, the feature amount is a multidimensional feature amount having a maximum of 64 × 3 = 192 dimensions. Based on the teacher data, the abnormal part detection unit 755 detects the processing unit partition as an abnormal part when the feature amount of the processing unit partition belongs to any of the abnormal part types A, B, and C. For example, a processing unit section having the feature amount P11 belonging to the abnormal part type B is detected as an abnormal part. Similarly, the processing unit section having the characteristic amount P13 belonging to the abnormal part type A and the abnormal part type C is detected as an abnormal part.

  Then, when the abnormal unit detection unit 755 detects the processing unit section as an abnormal unit, the abnormal unit detection unit 755 calculates a certainty factor in the detection of the abnormal unit. FIG. 10 is a diagram illustrating a certainty factor calculation method. As shown in FIG. 10, the certainty factor is calculated according to the distance from the center of the teacher data of the abnormal part type to which it belongs, for example. Here, based on the certainty factor calculated by the abnormal part detection unit 755, the abnormal part detection processing stop control unit 757 performs an abnormal part on the processing target image I (i) by the abnormal part detection unit 755 in step a15 of FIG. It is determined whether or not to stop the detection process. Specifically, the abnormal part detection processing stop control unit 757 compares this certainty factor with a reference certainty factor that is set in advance as a reference value for determining whether or not to stop the abnormal part detecting process. If it is above, the control which stops the abnormal part detection process about a unit division with the next highest priority will be performed, and the control which will transfer to the abnormal part detection process which made the next in-vivo image the process target image in time series will be performed. For example, if the reference certainty factor is “0.6”, the certainty factor in detecting the abnormal part is “0.6” or more, that is, if the feature amount of the processing unit section belongs to the region E21 in FIG. The abnormal part detection process for I (i) is stopped. In addition, when the teacher data is superimposed on each other as in the abnormal part type A and the abnormal part type C shown in FIG. The detection process stop control unit 757 performs comparison with the reference certainty factor using the certainty factor having a large value.

  The reference certainty factor may be set according to an external input by a user operation or the like. In this case, the control unit 770 functions as a reference certainty level setting unit. That is, the control unit 770 performs processing for displaying a notification of a setting request for the reference certainty factor on the display unit 730 and receives an input for determining the reference certainty factor via the input unit 720. FIG. 11 is a diagram illustrating an example of a notification screen for a reference certainty setting request. In the notification screen W10, an input box IB10 that accepts an input operation of the reference certainty factor together with a message that accepts an input of the reference certainty factor, an OK button BTN11 that finalizes the input operation and sets the reference certainty factor, and cancels the input operation A cancel button BTN13 is arranged. The user inputs a desired reference certainty value into the input box IB10 via the input unit 720, and presses an OK button BTN11 to set the reference certainty factor. For example, as shown in FIG. 11, when the reference certainty factor is set to “0.8” and the OK button BTN11 is pressed, the abnormal part detection processing stop control unit 757 has a certainty factor in detecting the abnormal part “0.8”. That is, in other words, when the feature amount of the processing unit section belongs to the area E23 of FIG. 10, the abnormal part detection process for the processing target image I (i) is stopped. In this way, if the setting value of the standard certainty can be changed on the notification screen of the setting request, the standard certainty can be set strictly or loosely as appropriate.

  Next, the image display process in step a27 in FIG. 5 will be described. FIG. 12 is a flowchart showing a detailed processing procedure of the image display processing. In this image display process, the image display processing unit 771 first initializes the image number i of the in-vivo image to be displayed to “i_start” (step d1). Subsequently, the image display processing unit 771 refers to the abnormal part detection data 745 and acquires the result of the abnormal part detection process for the in-vivo image I (i). The image display processing unit 771 detects an abnormal part in the in-vivo image I (i) when the abnormal part detection flag corresponding to the acquired image number i is set to “ON”. (Step d3: Yes). In this case, the image display processing unit 771 acquires the image data of the in-vivo image I (i) with the image number i (step d5), and obtains the abnormal portion detection result corresponding to the image number i acquired at step d3. Based on the identification number of the unit partition set as the processing unit partition, a unit partition that has not been subjected to the abnormal part detection process is detected. Then, when there is a unit partition that has not been subjected to the abnormal part detection process (step d7: Yes), the image display processing unit 771 controls the resumption of the abnormal part detection process with the unprocessed unit section as the processing unit section. (Step d9). That is, the image display processing unit 771 functions as a process reproduction control unit, and by the control here, the abnormal part detection unit 755 performs the abnormal part detection process for sequentially processing the unprocessed unit sections, and the in vivo image. An abnormal part can be detected for the entire area I (i). As a result, an image including an abnormal part can be detected at high speed from the entire time-series image, and an abnormal part can be detected in all areas in the image where the abnormal part exists by using a time when a person visually checks the image. It can be carried out.

  Then, the image display processing unit 771 determines the abnormal part region from the result of the abnormal part detection processing for each unit partition finally obtained, and displays the in-vivo image I (i) together with the abnormal part region on the display unit 730. A display process is performed (step d11). When a doctor or the like observes an in-vivo image, it is important from the viewpoint of observation support that the abnormal part to be noted is distinguished from the normal mucous membrane part, contents, and other parts that reflect bubbles. For example, the image display processing unit 771 sets, as an abnormal part region, a unit partition whose certainty obtained as a result of the abnormal part detection process is a predetermined value or more. Then, the image display processing unit 771 performs a process of displaying the in vivo image I (i) by distinguishing and displaying the abnormal part region from other regions. A threshold value used as a criterion for determining whether or not an abnormal region is to be used can be set as appropriate, and a reference certainty factor may be used, or a value larger or smaller than the reference certainty factor may be used.

  On the other hand, when “ON” is not set in the abnormal part detection flag corresponding to the image number i, the image display processing unit 771 determines that no abnormal part has been detected (No in step d3). Since the in-vivo image I (i) is not displayed, the process proceeds to step d13. It should be noted that an in-vivo image in which no abnormal part is detected may be displayed. For example, the display process may be performed in a shorter display time than the in-vivo image in which the abnormal part is detected.

  Then, the image display processing unit 771 increments the image number i to i = i + 1 (step d13), and determines the presence or absence of the in-vivo image I (i) to be displayed next by i = i_end. Then, when i ≦ i_end (step d15: Yes), the image processing unit 750 performs the processing from step d3 to step d13 again on the in-vivo image immediately after in time series. On the other hand, if i> i_end (step d15: No), the process returns to step a27 in FIG.

  As described above, according to the first embodiment, when the in-vivo images constituting the time-series image are processed and the position in the in-vivo image in which the abnormal part is reflected is detected, the processing target image is determined for each unit section. The priority order can be set for each unit section based on the abnormal part detection position detected in the previous in-vivo image in time series. Then, according to this priority, the unit section is sequentially processed as an abnormal unit detection process, and when the certainty level in the abnormal part detection obtained as a result of the abnormal part detection process is equal to or higher than the reference certainty level, the next priority order It is possible to stop the abnormal part detection process for the unit section having a high and move to the abnormal part detection process in which the next in-vivo images are processed in time series. Therefore, it is possible to suppress the processing related to the detection of the abnormal part to the minimum necessary, and to detect the in-vivo image including the abnormal part in a short time from the in-vivo images constituting the time series image.

(Embodiment 2)
Next, a second embodiment will be described. In the first embodiment, the abnormal part is detected using the DCT coefficient obtained during the expansion process of the processing target image. However, in the second embodiment, the abnormal part is detected using the pixel value information. The image processing apparatus according to the second embodiment is configured by replacing the image processing unit 750 with the image processing unit 760 shown in FIG. 13 in the configuration of the image processing apparatus 70 described with reference to FIG. 2 in the first embodiment. . As shown in FIG. 13, the image processing unit 760 constituting the image processing apparatus according to the second embodiment includes an area dividing unit 761, a feature amount calculating unit 763, a priority setting unit 765, and an abnormal part detecting unit 767. And an abnormal part detection process stop control part 769.

  Here, the flow of processing performed by the image processing apparatus according to the second embodiment will be described with reference to FIGS. In addition, the same code | symbol is attached | subjected to the process process similar to Embodiment 1. FIG.

  FIG. 14 is an overall flowchart illustrating a processing procedure performed by the image processing apparatus according to the second embodiment. In the second embodiment, in step e5, the region dividing unit 761 uses the restored in-vivo image I (i) as the processing target image I (i) to generate a region having a predetermined number of pixels (as in the first embodiment, “unit partition ”). Subsequently, the feature amount calculation unit 763 executes feature amount calculation processing (step e7), and the priority order setting unit 765 performs priority order setting processing (step e9).

  Next, the abnormal part detection unit 767 initializes the priority order k of the unit section to be processed to “1” (step e11), and sets the unit section of the priority order k in the processing target image I (i) as a processing unit. An abnormal part detection process is performed as a section (step e13). Then, the process proceeds to step e15, and the abnormal part detection process stop control unit 769 determines whether or not to stop the abnormal part detection process by the abnormal part detection unit 767 based on the result of the abnormal part detection process.

  First, the feature amount calculation process in step e7 in FIG. 14 will be described. FIG. 15 is a flowchart illustrating a detailed processing procedure of the feature amount calculation processing. In this feature amount calculation process, the feature amount calculation unit 763 performs the process of Loop A with each unit section as a processing target (step f1 to step f9).

  In the loop A, the feature amount calculation unit 763 first calculates an average value of the pixel values of the pixels in the unit partition as the feature amount of the unit partition to be processed (step f3). Then, the feature amount calculation unit 763 determines whether or not the unit partition to be processed is a mucosal region reflecting the mucous membrane based on the feature amount calculated in step f3. If it is a mucosal region (step f5: Yes), The unit section to be processed is set as a mucous membrane area (step f7). This is a process to be performed in consideration of whether or not the unit section is a mucosal region when the priority order setting section 765 sets the priority order for each unit section. For example, the feature amount is acquired as a sample based on the pixel value of the mucous membrane region in which the mucous membrane is projected in advance, and feature amount distribution information of the mucosal region is prepared as teacher data. Then, the feature quantity calculation unit 763 determines whether or not the processing target unit section is a mucosal region by comparing and collating the feature quantity calculated for the processing target unit section with the teacher data. It should be noted that, based on the pixel values of an area in which the area other than the mucous membrane has been projected in advance, for example, the contents area in which the contents are projected and the bubble area in which the bubbles are projected, the feature amount is obtained as a sample, Feature amount distribution information and the like are prepared as teacher data. Then, the feature quantity calculation unit 763 determines whether or not the processing target unit section is a content area or a bubble area by comparing and collating the feature quantity calculated for the processing target unit section with the teacher data. When it is neither an object area nor a bubble area, it may be determined as a mucosal area.

  If the processing of the loop A is executed for all the unit partitions in accordance with the above procedure, the process returns to step e7 in FIG. 14, and then proceeds to step e9. Information on whether or not each unit section obtained as a result of the feature amount calculation processing is a mucosal region is temporarily stored in the storage unit, and is referred to during priority setting processing.

  Next, the priority setting process in step e9 in FIG. 14 will be described. FIG. 16 is a flowchart showing a detailed processing procedure of priority order setting processing. In the second embodiment, the priority order setting unit 765 sets the priority order in step b5 or step b7, and then determines whether or not the unit partition is a mucosal region determined in the feature amount calculation process described with reference to FIG. The priorities are corrected by weighting (step g9). In the in-vivo image showing the inside of the body cavity, the observer wants to confirm the mucosal region where the mucous membrane is mainly reflected. This is because there is a possibility that an abnormal part appears in the mucous membrane region. In the first embodiment, the unit compartment reflecting the mucous membrane region is identified as a unit compartment that may correspond to the abnormal part, and this is designated as the unit compartment. Add priority to correct priority. Specifically, the priority order setting unit 765 increases the priority order of unit sections that are mucosal areas, and lowers the priority order of areas that are not mucosal areas to correct the priority order.

  Next, the abnormal part detection process by step e13 of FIG. 14 is demonstrated. FIG. 17 is a flowchart showing a detailed processing procedure of the abnormal part detection processing. Here, the region in which the abnormal part appears is considered to be a portion where a change in the pixel value is seen from the pixels constituting the normal mucosal region existing around. Therefore, in this abnormal part detection process, a process of calculating a pixel value change amount using one pixel in the processing unit section as a target pixel is performed for all pixels in the processing unit section, and abnormal part candidate pixels are detected. In calculating the pixel value change amount, a color component corresponding to the blood absorption band is used. When the in-vivo image is an RGB image, a pixel value change due to a lesion tends to appear in a G component that is close to the blood absorption band and has a relatively high sensitivity and resolution. The pixel value change amount is calculated as a feature amount using the G component image of the RGB image. That is, the abnormal part detection unit 767 first performs each process of loop B by sequentially setting each pixel constituting the processing unit section as a target pixel (step h1 to step h5).

In the loop B, the abnormal part detection unit 767 first calculates a pixel value change amount of the pixel of interest (step h2). FIG. 18 is an explanatory diagram illustrating a method for calculating a pixel value change amount. The calculation of the pixel value change amount is performed in, for example, four directions including a horizontal direction, a vertical direction, a right-up diagonal direction, and a right-down diagonal direction around the target pixel IP which is one pixel in the processing unit section E30. That is, as shown in FIG. 18, each pixel IA _hor , IB _{hor of} the unit block separated by a predetermined length λ or a length corresponding to the pixel number λ in the horizontal direction from the target pixel IP, Each pixel IA _ver , IB _ver , each pixel IA _sla , IB _sla in the unit partition separated in the upward and diagonal direction, and each pixel IA _bac and IB _bac in the unit partition separated in the downward and diagonal direction are the surrounding pixels. Based on the pixels and their surrounding pixels, a pixel value change amount is calculated.

The pixel value change amount V _{hor in} the horizontal direction of the _target pixel IP is expressed by the following equation (1) from the _target pixel IP and the surrounding pixels IA _hor and IB _{hor in} the horizontal direction.

The pixel value change amount V _ver in the vertical direction of the target pixel IP is expressed by the following equation (2) from the target pixel IP and the surrounding pixels IA _ver and IB _{ver in} the vertical direction.

The pixel value change amount V _{sla in the} diagonally upward direction of the _target pixel IP is expressed by the following equation (3) from the _target pixel IP and the surrounding pixels IA _sla and IB _{sla in the} diagonally upward direction.

The pixel value change amount V _{bac in the} diagonally downward direction of the _target pixel IP is expressed by the following equation (4) from the _target pixel IP and the surrounding pixels IA _bac and IB _{bac in the} diagonally downward direction.

Here, the pixel value change amounts V _hor , V _ver , V _sla , and V _bac calculated according to the equations (1) to (4) are comprehensively described as a pixel value change amount V _dir . Here, the subscript dir is a subscript indicating a predetermined direction (in this embodiment, any of horizontal, vertical, right-up diagonal, and right-down diagonal) centered on the pixel of interest IP.

When the region of the target pixel IP is more convex than the surroundings, the pixel value change amount V _dir takes a positive value. On the other hand, when the region of the pixel of interest IP is more concave than the surroundings, the pixel value change amount V _dir takes a negative value. This is because, in general, a region that is more convex than the surroundings is brighter and has a higher luminance value than the surroundings, and a region that is recessed more than the surroundings is darker and has a lower luminance value.

Subsequently, as shown in FIG. 17, the abnormal part detection unit 767 determines whether or not a significant change is observed in the pixel value between the _target pixel and each surrounding pixel based on the calculated pixel value change amount V _dir. If it is determined and a change is observed (step h3: Yes), the target pixel is set as an abnormal part candidate pixel (step h4). Specifically, the abnormal part detection unit 767, based on the pixel value change amount calculated for each direction, when the pixel value change amount in all directions is larger than a predetermined convex threshold (ConvexityTh), that is, V _hor > If ConvexityTh and V _ver > ConvexityTh and V _sla > ConvexityTh and V _bac > ConvexityTh, it is determined that the target pixel is convex with respect to the surroundings, and is detected as an abnormal part candidate pixel. On the other hand, the abnormal part detection unit 767, based on the pixel value change amount calculated for each direction, when the pixel value change amount in all directions is smaller than a predetermined concave threshold (ConcaveTh), that is, V _hor <ConcaveTh and V _{If ver} <ConcaveTh and V _sla <ConcaveTh and V _bac <ConcaveTh, it is determined that the target pixel is concave with respect to the surroundings, and is detected as an abnormal part candidate pixel. Then, the abnormal part detection unit 767 performs the same process using all pixels in the processing unit section as the target pixel, detects abnormal part candidate pixels in the processing unit section, and ends the loop B. Thereafter, the abnormal part detection unit 767 calculates the area of the abnormal part candidate pixel, the average pixel value information of the R, G, and B components in the abnormal part candidate pixel, the pixel value variance, and the like as the feature amount of the abnormal part candidate pixel, Based on this value, it is determined whether the processing unit section is an abnormal part (step h6). Regarding the determination method using the feature amount, the method using the teacher data already described can be used.

  And it returns to step e13 of FIG. 14, and transfers to step e15 after that. In step e15, the abnormal part detection process stop control unit 769 determines whether or not to stop the abnormal part detection process for the processing target image I (i) by the abnormal part detection unit 767. Specifically, the abnormal part detection process stop control unit 769 stops the abnormal part detection process for the unit partition having the next highest priority when detecting a processing unit partition as an abnormal part as a result of the abnormal part detection process. Control is performed, and control is performed to shift to the abnormal portion detection processing in which the next in-vivo images are processed in time series.

  As described above, according to the second embodiment, the same effects as those of the first embodiment can be obtained, the processing for detecting an abnormal part can be suppressed to the minimum necessary, and the in-vivo images constituting the time-series image can be selected. The in vivo image including the abnormal part can be detected in a short time.

  As described above, the two preferred embodiments of the present invention have been described. However, the present invention is not limited to the above-described embodiments, and various modifications can be made without departing from the spirit of the invention.

  For example, in the above-described first embodiment, the degree of certainty in abnormal part detection is calculated, and the case where the abnormal part detection process is stopped based on the calculated degree of certainty has been described. However, the degree of importance of the detected abnormal part is calculated. Then, the abnormal part detection process may be stopped based on the importance. For example, an importance level table in which importance levels are set for the abnormal part types A, B, and C described with reference to FIG. 19 is stored in the storage unit. FIG. 19 is a diagram illustrating a data configuration example of the importance level table. As shown in FIG. 19, in the importance level table, the importance value is set in association with the abnormal part type. In this case, the abnormal part detection unit calculates the importance of the abnormal part detected according to the importance table. At this time, when the teacher data is superimposed on each other like the abnormal part type A and the abnormal part type C shown in FIG. The part detection unit selects an abnormal part type that is close to the center and calculates the importance. Then, the abnormal part detection process stop control unit compares this importance with a reference importance set in advance as a reference value for determining whether or not to stop the abnormal part detection process. Part. For example, if the reference importance is “0.5”, the feature amount of the processing unit section is an abnormal part type (abnormal part type A) in which a value of “0.5” or more is set as the importance in the importance table. Or abnormal part type B), the abnormal part detection process for the next highest priority unit section is stopped, and the process proceeds to the abnormal part detection process using the next in-vivo image as the processing target image in time series. Is done.

  The reference importance may be set according to an external input by a user operation or the like. In this case, the control unit 770 functions as a reference importance setting unit. That is, the control unit 770 performs processing for displaying a notification of a reference importance setting request on the display unit 730 and receives an input for determining the reference importance via the input unit 720. FIG. 20 is a diagram illustrating an example of a notification screen for a reference importance setting request. The notification screen W20 includes a message indicating that the input of the reference importance is accepted, an input box IB20 that accepts an input operation of the reference importance, an OK button BTN21 that confirms the input operation and sets the reference importance, and cancels the input operation. A cancel button BTN23 is arranged. The user inputs a desired reference importance value into the input box IB20 via the input unit 720, and presses an OK button BTN21 to set the reference importance. As described above, if the setting value of the reference importance can be changed on the notification screen of the setting request, the reference importance can be set strictly or appropriately according to the importance of the abnormal part to be detected. .

Moreover, it is good also as detecting an abnormal part using both the value of reliability and the value of importance. FIG. 21 is a diagram for explaining a method for detecting an abnormal part in this case, and FIG. 21 shows teacher data for each of the abnormal part types E, F, and G. Here, for example, the importance of the abnormal part type E is “0.2”, the importance of the abnormal part type F is “0.8”, and the importance of the abnormal part type G is “0.9”. Suppose there is. Further, it is assumed that the standard certainty is “0.5” and the standard importance is “0.6”. For example, as a result of the abnormal part detection processing being performed by the abnormal part detection unit, a feature amount P21 belonging to a region where the teacher data of the abnormal part types E, F, and G shown in FIG. part type E is calculated as "0.1" in accordance with the distance from the center O _E for, it is calculated as "0.7" for abnormality type F according to the distance from the center O _F, from the center O _G abnormality unit type G In the case where “0.3” is calculated according to the distance, the importance degree of the abnormal part type F having the certainty degree “0.5” or more is “0.8”, so the next highest priority is given. The abnormal part detection process for the unit section is stopped, and the process proceeds to an abnormal part detection process in which the next in-vivo images are processed in time series. That is, the feature quantity of the section to be processed belongs to the teacher data of the abnormal part type F or abnormal part type G whose importance is equal to or higher than the reference importance, and the certainty in detecting the abnormal part obtained from this feature quantity is “0” .5 ”or more, that is, when the feature amount of the processing unit section belongs to the area E41 or the area E43 indicated by the broken line in FIG. 21, the abnormal part detection process is stopped.

  Further, in the above-described first embodiment, the case where the abnormal part of the unit section divided into regions using the DCT coefficient is detected and the pixel value change amount is detected in the second embodiment has been described. However, the present invention is not limited to these, and it can be appropriately detected using a known technique. For example, the luminance, chromaticity, color difference, etc. of the pixels constituting each unit section are used, and the color information of each unit section is changed to color information of abnormal parts such as bleeding, color information of normal mucous membrane areas, contents and bubble areas. The abnormal part may be detected by comparing with a reference value based on the color information.

  In the priority order setting process, the priority order may be set in consideration of the certainty factor obtained as a result of the abnormal part detection process performed on the previous (for example, immediately preceding) in-vivo image in time series. FIG. 22 is a diagram for explaining the priority setting method in this case, and the confidence of each unit section obtained as a result of the abnormal part detection processing performed on the in-vivo image immediately before the processing target image in time series. Shows the degree. Here, based on the certainty factor “0.8” obtained for the unit section E51 as a result of performing the abnormal part detection process on the immediately preceding in-vivo image in time series, the unit section E51 is detected as the abnormal part detection position. Is recorded. In this case, in the next in-vivo image in chronological order, the priority order of the unit section E51 is set to “1”, and for the other unit sections, the unit section closer to the unit section E51 is higher. Priorities are set, and at this time, the priorities may be corrected based on the certainty factor calculated for the unit section subjected to the abnormal part detection process. Specifically, among the unit sections for which the certainty factor is obtained by performing the abnormal part detection process in the immediately preceding in-vivo image in the time series shown in FIG. It is good also as correcting the priority of each unit partition so that the priority of the unit partition E53 from which the certainty factor was obtained, and the unit partition E55 becomes high. According to this, taking into account the certainty factor obtained as a result of the abnormal part detection processing performed on the previous (for example, immediately preceding) in-vivo image in time series, the priority order of the unit sections obtained with high certainty factor is determined. Can be set high.

  Further, in the priority setting process, when obtaining the importance for the abnormal part detected in the abnormal part detection process, the result of the abnormal part detection process performed on the previous (for example, immediately preceding) in-vivo image in time series is obtained. The priority can be set in consideration of the importance level. For example, among the importance calculated for the unit section that has been subjected to the abnormal part detection process in the immediately preceding in-vivo image in time series, the unit that has obtained a relatively large importance that is less than the reference importance The priority order of each unit section may be modified so that the priority order of the sections becomes higher. According to this, in consideration of the importance obtained as a result of the abnormal part detection processing performed on the previous in vivo image in time series (for example, immediately before), the priority order of the unit sections obtained with high importance is determined. Can be set high.

  Moreover, although the abnormal part detection process was performed for every unit division, it is good also as performing an abnormal part detection process for every pixel. For example, in the second embodiment, the image is divided into unit partitions, and the pixel value change amount is calculated for each unit partition to detect the abnormal portion. As each pixel of interest, a pixel value change amount is calculated based on the pixel of interest and a predetermined number of pixels λ or surrounding pixels located at a distance corresponding to the number of pixels λ, and compared with the surrounding pixels. It is also possible to detect a target pixel in which a significant change in value is detected as an abnormal part. In this case, the priority order is set for each pixel based on the pixel at the abnormal part detection position detected in the in-vivo image that is close in time series, and each pixel is sequentially set as the target pixel according to this priority order. An abnormal part detection process is performed.

  In each of the above-described embodiments, the priority order is set for each unit section based on the abnormal part detection position detected in the previous (for example, immediately preceding) in-vivo image in time series. When the in-vivo images in the chronological order are processed first, priority is given to each unit section based on the abnormal part detection position detected in the subsequent in-vivo image in the time series (for example, immediately after). It is possible to set.

  Further, in each of the above-described embodiments, when no abnormal portion is detected in the previously processed in-vivo image, priority is given to each unit section according to the priority order set for the previously processed in-vivo image. Although it was decided to set, the priority of the unit sections existing at a specific location, for example, the center and the periphery of the image to be processed, is set to the highest priority, and the priority order of each unit partition is set so as to decrease as the distance from the location increases It is good to do. Similarly, when processing the first in-vivo image in time series, for example, when processing first, similarly set the highest priority order of unit sections existing in a specific place, for example, the center and the periphery of the processing target image, It is good also as setting the priority of each unit division so that it may become so low that it leaves | separates from a place. If abnormal part detection processing is to be performed for each pixel, the priority order of pixels existing in a specific place, for example, the center and the periphery of the processing target image is set to the highest, and the priority is lowered as the distance from the place increases. The priority order of each pixel may be set. When projecting inside a tubular living body, there may be abnormalities that are easier to recognize in the surroundings than in the center where the back of the tube appears dark.In that case, the priority is set spirally from the periphery of the image toward the center. It is possible to inspect more efficiently when set.

  Alternatively, the image processing device may perform processing for detecting an abnormal portion reflected in the time-series image, and the time-series image in which the abnormal portion is detected on another display device may be displayed together with the abnormal portion region.

1 is a schematic diagram illustrating an overall configuration of an image processing system including an image processing apparatus. It is a block diagram explaining the functional structure of an image processing apparatus. It is a figure which shows the DCT coefficient acquired for every unit division in the case of an expansion | extension process. It is a figure which shows the data structural example of abnormal part detection data. 3 is an overall flowchart illustrating a processing procedure performed by the image processing apparatus according to the first embodiment. 4 is a flowchart illustrating a detailed processing procedure of priority order setting processing according to the first embodiment. It is explanatory drawing for demonstrating the setting method of a priority. It is explanatory drawing for demonstrating the setting method of a priority. 3 is a flowchart illustrating a detailed processing procedure of abnormal part detection processing according to the first embodiment. It is a figure which shows an example of the teacher data of the feature-value of an abnormal part. It is a figure explaining the calculation method of reliability. It is a figure which shows an example of the notification screen of the setting request of reference | standard reliability. It is a flowchart which shows the detailed process sequence of an image display process. 6 is a diagram illustrating a configuration of an image processing unit of an image processing apparatus according to a second embodiment. FIG. 10 is an overall flowchart illustrating a processing procedure performed by the image processing apparatus according to the second embodiment. 10 is a flowchart illustrating a detailed processing procedure of a feature amount calculation process according to the second embodiment. 10 is a flowchart illustrating a detailed processing procedure of priority order setting processing according to the second embodiment. 10 is a flowchart illustrating a detailed processing procedure of abnormal part detection processing according to the second embodiment. It is explanatory drawing explaining the calculation method of pixel value variation | change_quantity. It is a figure which shows the data structural example of an importance table. It is a figure which shows an example of the notification screen of the setting request | requirement of a reference | standard importance. It is a figure for demonstrating the detection method of the abnormal part in a modification. It is a figure explaining the setting method of the priority in the modification.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Capsule-type endoscope 30 Receiving device A1-An Receiving antenna 50 Portable recording medium 70 Image processing device 710 Image acquisition part 720 Input part 730 Display part 740 Storage part 741 Image processing program 743 Priority order setting data 745 Abnormal part detection data 750 Image processing unit 753 Priority order setting unit 755 Abnormal part detection unit 757 Abnormal part detection processing stop control unit 770 Control unit 771 Image display processing unit 1 Subject

Claims (23)

An image processing apparatus that processes a time series image composed of a plurality of images taken in time series,
Priority order setting means for setting a priority order for pixels or regions to be processed in the images constituting the time-series images;
A detection process for detecting an area corresponding to an abnormal part of a subject in order from a pixel or area having a high priority set by the priority setting means is performed on a plurality of images constituting the time-series image. An abnormal part detecting means for sequentially performing
A process stop control means for controlling the stop of the detection process in each image by the abnormal part detection means based on the result of the detection process;
Detection result recording means for recording a result of the detection process;
An image processing apparatus comprising:   The priority order setting means sets the priority order in the processing target image based on a result of the detection processing in a neighboring image that is close in time series to the processing target image that is a target of the detection processing. The image processing apparatus according to claim 1, wherein the image processing apparatus is set.   The priority order setting means sets a higher priority order in the processing target image for a pixel or region existing at a position corresponding to an abnormal part detection position where an abnormal part is detected in the neighboring image. 2. The image processing apparatus according to 2.   4. The image according to claim 3, wherein the priority order setting unit sets a higher priority order in the processing target image for a pixel or region closer to a position corresponding to the abnormal portion detection position in the neighboring image. Processing equipment.   The priority order setting means has feature amount calculation means for calculating a feature amount of each pixel or each region in a plurality of images constituting the time-series image, and sets the priority order based on the feature amount. The image processing apparatus according to claim 1, wherein the image processing apparatus is an image processing apparatus.   The priority order setting means specifies a pixel or region that is highly likely to correspond to the abnormal portion of the subject based on the feature amount, and sets the priority order of the specified pixel or region high. The image processing apparatus according to claim 5.   The process stop control means stops the detection process performed by the abnormal part detection means for a pixel or region having the next highest priority according to a result of the detection process for the pixel or region having a higher priority, The image processing apparatus according to claim 1, wherein control is performed to shift to the detection process for an image for which the detection process has not been processed.   The image processing apparatus according to claim 1, wherein the abnormal part detection unit calculates a certainty factor that is an index of reliability of the detection process.   The image processing apparatus according to claim 8, wherein the process stop control unit controls stop of the process by the abnormal part detection unit based on the certainty factor.   10. The method according to claim 9, further comprising a reference certainty factor setting unit that sets a reference certainty factor as a criterion for determining whether or not the processing stop control unit stops the processing by the abnormal part detecting unit according to an external input. The image processing apparatus described.   9. The priority order setting unit sets a higher priority order of pixels or regions existing at positions where high certainty is obtained in the detection process in an image that is close in time series. 10. The image processing device according to any one of 10.   The image processing apparatus according to claim 1, wherein the abnormal part detection unit calculates importance for the detected abnormal part.   The image processing apparatus according to claim 12, wherein the process stop control unit controls the stop of the detection process based on the importance of the abnormal part obtained during the detection process.   The image processing according to claim 13, further comprising reference importance level setting means for setting a reference importance level as a criterion for determining whether or not the processing stop control means stops the detection process according to an external input. apparatus.   13. The priority order setting unit sets a high priority order of pixels or regions existing at positions where high importance is obtained when detecting an abnormal part in an image that is close to a time series. 14. The image processing device according to any one of 14.   In the case where the image to be subjected to the detection process is an image to be processed first, or the region corresponding to the abnormal part is not detected in the image that is close in time series, the priority order setting means The image processing apparatus according to claim 1, wherein a higher priority is set for a pixel or a region existing in a specific location of the processing target image that is the target of the image processing. Priority order recording means for recording the setting result of the priority order by the priority order setting means,
The priority order setting unit sets the priority order for a pixel or a region to be processed based on a priority order set for neighboring images in time series. The image processing apparatus according to any one of 16.   18. The display processing unit according to claim 1, further comprising a display processing unit configured to display an image constituting the time-series image on a display unit together with a result of the detection process for the image. The image processing apparatus described.   The display processing unit includes a process resumption control unit that restarts the detection process and detects an abnormal part for a pixel or region that is not subjected to the process of detecting the abnormal part. The image processing apparatus according to claim 18, wherein a process for displaying the image is performed.   The image processing apparatus according to claim 1, wherein the images constituting the time-series images are images compressed using DCT encoding.   21. The image processing apparatus according to claim 20, wherein the area is a unit section of processing at the time of DCT encoding.   The image processing apparatus according to claim 20 or 21, wherein the abnormal part detecting unit detects an abnormal part based on a DCT coefficient calculated at the time of DCT encoding. To a computer that processes a time-series image composed of a plurality of images taken in time series,
A priority setting step for setting a priority for a pixel or a region to be processed in an image constituting the time-series image;
Detection processing for detecting an area corresponding to an abnormal part of a subject in order from a pixel or area having a high priority set in the priority setting step is performed on a plurality of images constituting the time-series image. Abnormal part detection step sequentially performed,
A process stop control step for controlling the stop of the detection process in each image by the abnormal part detection step based on the result of the detection process;
A detection result recording step for recording a result of the detection process;
An image processing program for executing

Images