JPH03174670A - Image processor - Google Patents

Abstract

PURPOSE:To simply and quickly detect the boundary of an optional subject included in a general image and an area inside or outside the boundary by adding a label to each picture element, selecting a noted picture element out of a pointed area, updating the label based on the characteristic of the noted picture element, deciding whether the set end conditions are satisfied or not, and selecting a picture element that satisfies the specific conditions out of the label picture elements with which the processing is through. CONSTITUTION:An operator encloses a subject 52 with a closed curve via an input part 33 to designate a pointed area 53 and adds the labels to each picture element included in the area 53 via a label adding means 13. Then the operator selects a noted picture element via a noted picture element selection means 14 and updates the label value of the noted picture element via a label value updating means 15. An end deciding means 16 decides whether the ending conditions of processing are satisfied or not. Then a subject mask is produced via a picture element detection means 17. Thus it is possible to quickly detect the boundary of an optional subject included in a general image and an area inside or outside the boundary with a simple operation.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to an image processing device that detects the boundaries of an arbitrary object in an image such as a digitally expressed photograph or video image. Based on this information, it is possible to obtain a contour image of the object or obtain a mask image of the object, which can be used for various image processing such as cropping, deforming, and correcting/changing the color of the object.

(Prior Art) The operation of cutting out an object that is part of a digitally expressed photograph or video image is an important operation that is frequently used when performing image editing.

Conventional methods include enlarging the target area and cutting out the area by tracing the surrounding area using straight lines, curved lines, etc. However, this method is extremely time-consuming when cutting out large or complex objects.

Therefore, as a method to reduce this effort, there is a method of specifying the range of colors that exist in the target (for example,
8968 (see Japanese Patent Application Laid-open No. 8968), and one in which edges are extracted by micro-edge processing (for example, see Japanese Patent Laid-Open No. 62-219179).

(Problem to be solved by the invention) However, although the former method is effective for objects that are composed of fairly uniform colors, there are variations in color within the object due to light reflection, texture, etc. Not applicable to general subjects.

Furthermore, in the latter method, the extracted edges often do not become closed curves, and noise and texture are also extracted, so very difficult post-processing is required.

The present invention solves the shortcomings of the conventional technology, and easily and quickly detects the boundary or mask image of an arbitrarily specified target image from a general image that contains light reflection, texture, noise, etc. The object of the present invention is to provide an image processing device that can perform the following steps.

(Means for Solving the Problems) As shown in FIG. A pixel of interest selection means 14 selects a pixel from the region, and a label value (62) and a feature (for example, pixel value [74,
113]) and a label value changing means 15 for changing the label value of the pixel of interest based on the label value and characteristics of a group of pixels in the vicinity of the pixel of interest; and a pixel detection means 17 that detects pixels having a label that satisfies a specific condition from the label image as pixels belonging to the boundary of the object, inside or outside the object.

(Function) The basic flow of the present invention is shown in FIG. The basic operation of the present invention will be briefly explained below.

(2) A label is attached to each pixel by the label attaching means 13. There are no particular restrictions on how label values are assigned to pixels in the designated area. Regarding the labels of pixels outside the designated area, there is a difference in label value between pixels outside and inside the designated area, and there is not much difference in label values between pixels on the outside or between pixels inside the designated area. Just do it.

For example, it is provided as shown in FIG. 6(a).

Further, the label can be attached only to the designated area, and in this case, the label only needs to have a difference in value between the outside and inside of the designated area, and there are no particular restrictions on the value or step. For example, as shown in FIG. 10(a), it is applied so that it increases monotonically toward the inside.

(2) Using the pixel of interest selection means 14, a pixel (hereinafter referred to as a pixel of interest) to be processed in step (2) is arbitrarily selected from within the designated area.

■ Features (pixel value) of the pixel of interest selected in step ■
, the label value and the characteristics (pixel value) of its neighboring pixels. Based on the label value, the label value changing means 15 updates the label of the pixel of interest. This updating process is performed so that the label value of the pixel of interest becomes close to the label value of a neighboring pixel having characteristics similar to the pixel of interest.

(2) The termination determining means 16 determines whether preset conditions for termination are satisfied. If it is not satisfied, move to step ■.

(2) By using the pixel detection means 17 to detect pixels that satisfy a specific condition from the processed label image, the boundary of the object is detected or a mask is created.

In FIG. 1, a control section 11 controls the entire apparatus, and a storage section 12 stores images, label values, and the like.

(Examples) Hereinafter, the present invention will be explained in further detail using two examples.

First Embodiment FIG. 3 is a block diagram of an image cutout processing apparatus as an embodiment of the present invention, and shows an example in which a label is attached both inside and outside the indicated area.

This device includes a control section 11 that controls the entire device, a storage section 12 that stores images, label values, etc., a mask creation section 31 that creates a cutout mask, a display section 82 that displays a target image, and an operator. Input section 3 for giving instructions
Equipped with 3. The mask creation section 31, which is a feature of this embodiment, is configured by the labeling means 11 described in FIG.
, a pixel of interest selection means 12, a label value change means 15, an end determination means 16, and a pixel detection means 17.

Hereinafter, this first step will be performed according to the flowchart shown in FIG.
The operation of the embodiment will be explained.

[1] Display the target image 51 on the display unit 32.

The displayed image 51 at this point is shown in FIG. 5(a).

In this example, a lemon is drawn as the target 52 to be cut out.

[2] The operator uses the input unit 33 to designate the designated area 53 by surrounding the object 52 with a closed curve having a width including the boundary. The displayed image at this point is shown in FIG. 5(b).

[3] The labeling means 13 gives a label to each pixel in the designated area 5S. This label is in instruction area 5
Different values are assigned to pixels inside and outside 3. Here, as an example, it is assumed that the information is given as shown in FIG. 6(a).

[4] Using the pixel of interest selection means 14, arbitrarily select a pixel of interest from within the designated area using, for example, a random number.

[5] The label value of the pixel of interest selected in step [4] is updated by the label value changing means 15. An example of how to obtain a new label value is shown below.

Appropriate pixels are selected from around the pixel of interest and are used as reference pixels. In the n-dimensional feature space created by the pixel values, the reference pixel closest to the pixel of interest is found, and the average of the label values of that pixel and the pixel of interest is set as the new label value. In addition,
The average may be a weighted average, in which case the weight of the pixel of interest may be set to O.

In Figure 7, the four pixels in the four corners of the 5x5 pixels in the vicinity of the pixel of interest are taken as reference pixels (k = 4).
, an example of processing using the above method when n=1 is shown. When n=1, the distance in the feature space can be defined by the difference in pixel values, so in FIG. 7, the pixel 73 is the closest among the reference pixels 72. The label value of pixel 78 is 6, and the label value of pixel of interest 71 is 2, so the new label value is (8+2
)/2=4.

Through the above processing, the label values are changed so that the label values of pixels having similar characteristics become similar. Note that the method of changing the label value is just one example, and as will be described later with reference to FIG. 11, the label value of the pixel closest to the pixel of interest among the neighboring pixels may be set as the updated label value of the pixel of interest.

Furthermore, the label may be updated by selecting not only the pixel closest to the pixel of interest, but also a plurality of pixels in descending order of their proximity.

Since there is no limit to the number n of pixel features used in this step, features such as the dispersion, differential value, and gradient direction of pixel values in addition to the Buddhist coordinates can be used to create images with changes in texture and color. It is possible to improve responsiveness.

[6] The end determination means 16 determines step [4]
It is determined whether the condition for terminating the repeat process in [5] is met. For example, the termination condition may be a method of displaying the progress and waiting for the operator's termination instruction, or a method of
One possible method is to end the process after the label change process is completed.

[7] Using the pixel detection means 17, create a mask of the target. Through the processing up to Step 6, the label image changes as shown in FIG. 6(b) to FIG. 6(C). Therefore, for example, by binarizing the label image using a threshold specified by the operator, the boundary and areas inside and outside the boundary can be detected, and thereby a mask image of the target can be created.

[8] Cut out the desired object by multiplying the mask and the original image.

As described above, according to the first embodiment, it is possible to easily manipulate the boundaries of arbitrary objects in general images that have color changes and textures, which were conventionally common, and areas within or outside the boundaries. can be detected quickly. Therefore, it becomes easy to create a mask image, and various image processing such as cropping and modification of the target image can be easily performed.

Furthermore, according to the first embodiment, since the label change process is performed using information on pixels other than the designated area, the mask can be created with higher precision and faster than the second embodiment described below. Can be done.

Second Embodiment FIG. 8 is a block diagram of a second embodiment of the present invention, and shows an example in which a label is attached only to a designated area. This device includes a control section 11 that controls the entire device, a storage section 12 that stores images, label values, etc., a boundary detection section 8, a display section 82 that displays a target image, and an operator that gives instructions. and a mask creation unit 84 that creates a cutout mask from the detected boundaries.

A boundary detection unit 81 that is a feature of this second embodiment
is composed of the labeling means 13, the target pixel selecting means 14, the label value changing means 15, the end determining means 16, and the pixel detecting means 17 described in FIG.

1. The operation of this second embodiment will be explained below along the flowchart shown in FIG.

[1] Display the target image 51 on the display unit 82.

The displayed image 51 at this point is shown in FIG. 5(a).

[2] The operator uses the input unit 83 to designate the designated area 53 by surrounding the object 52 with a closed curve whose width includes the boundary. The displayed image at this point is shown in FIG. 5(b).

[3] The labeling means 13 gives a label to each pixel in the designated area 53. This label is displayed in the display area 53.
Different values are assigned to pixels close to the inside and pixels close to the outside. Here - as an example, Figure 10(a)
As shown in (b), label values are assigned so that they become larger toward the inside. As an example of a labeling method, a certain label is assigned to the pixels located at the outermost position of the instruction area 53 while tracing them along a closed curve, and all labels are assigned to the pixels located at the outermost position. Then, in the same way, the pixel located one step inside the outermost one is traced along the closed curve, changing its value and giving it a 2-label. Instruct area 5 to perform such processing.
Repeat until all pixels in step 3 have been labeled.

[4] Using the pixel of interest selection means 14, arbitrarily select a pixel of interest from within the designated area using, for example, a random number. However, the outermost and innermost pixels of the instruction area 53 (indication area 53
The pixel 63 (which is in contact with other areas) is excluded from the selection target.

[5] The label value of the pixel of interest selected in step [4] is updated by the label value changing means 15. This updating method is the same as in the first embodiment, and an example will be explained here.

Among the kXI neighboring pixels of the pixel of interest located in the designated area, find the pixel closest to the pixel of interest in the n-dimensional feature space created by the pixel values of each pixel, and use the label value of that pixel as a new label of the pixel of interest. value.

In FIG. 11, k=1=3. An example of processing using the two-note system when n=1 is shown below. When n = 1, the distance in the feature space can be defined by the difference in pixel values, so in Fig. 11, the pixel of interest 1
The pixel 112 is the closest pixel among the 11 3×3 neighboring pixels. Since the label value of pixel II2 is 4, the new label value of the pixel of interest 111 is also updated to 4.

Through the above processing, the label values are changed so that the label values of pixels having similar characteristics become similar.

There is no limit to the number n of pixel features used in this step, so features such as dispersion, differential value, and gradient direction of pixel values are used in addition to color standard values to create images with changes in texture and color. can improve responsiveness.

Further, in the above example, l is described as fixed for convenience of explanation, but it is not necessary to be fixed during processing. For example, ・Change I depending on a random number.

- Increase l until a certain condition is met.

It is also possible to change the size of the neighborhood region using methods such as this to speed up the label change. Furthermore, by appropriately thinning out pixels without using all the pixels in the k×1 area, the number of calculations can be reduced and the processing speed can be increased.

[6 pieces] The end determination means 16 determines that step [4 pieces [
It is determined whether the condition for terminating the 5-item repeat process is satisfied.

As in the first embodiment, the termination conditions may include, for example, a method in which the progress is displayed and waits for an operator's instruction to terminate, or a method in which the update process ends when the amount of change in the label is less than or equal to a threshold value continues for m or more consecutive times. is possible.

[7] Use the pixel detection means 17 to find the boundary of the object. Through the processing up to Step 6, the label image changes as shown in FIG. 10(b) to FIG. 10(C). Therefore, for example, the position 104 where the target boundary exists can be detected by a method of detecting a point where a label value intersects a threshold value specified by an operator, or by a method of tracing a line where the label value changes rapidly.

[8] The operator uses the input unit 88 to give an instruction as to whether to cut out the inside or outside of the boundary.

15-[9] According to the instructions in step [7], the mask creation unit 84 creates a cutout mask 54 in which the inside or outside of the boundary is filled. The processing results at this point are shown in FIG. 5(C).

[10] Cut out the desired object by multiplying the mask and the original image.

As described above, according to the second embodiment, it is possible to quickly detect the boundary of an arbitrary object in a general image containing color changes, textures, etc., which has been difficult in the past, with a simple operation. Therefore, by creating a contour line or a mask image from the detected boundary of the object, various image processing such as cropping and modification of the object image can be easily performed.

In the first and second embodiments described above, all processing is performed in pixel units. By doing so, you can further increase the processing speed.

An example is shown below.

As shown in Fig. 12 (a) and (b), pxq pixels - 6 (in the case of Fig. 12, p = q = 2) are combined into a unit area 1.
22, and in the first step, processing is performed by regarding this unit area 22 as a pixel. At this time, as the feature amount of the unit area 122, the average value of the feature amount of pixels within the area is used. As a result, a unit area 123 including the boundary is found as shown in FIG. 12(C), and that portion is slightly expanded to form a new designated area 124 as shown in FIGS. 12(d) and (e). In the second step, pixel-by-pixel processing is performed on the new designated area 124. This method increases the processing to two stages, but since the number of elements to be processed at each stage is small,
Overall, the processing speed can be improved.

Further, for example, if pixels having the same pixel value as adjacent pixels are grouped together as a small area and processing is performed on a unit of this small area, it is also possible to directly determine the boundary without the second stage of pixel-by-pixel processing.

(Effects of the Invention) According to the present invention, it is possible to easily and quickly detect the boundary of an arbitrarily specified target image and create a mask image from a general image containing light reflection, texture, noise, etc. be able to.

Furthermore, by performing label change processing using information on pixels other than the designated area, it is possible to create a mask for a general image that includes color changes, textures, etc. with higher accuracy and speed.

[Brief explanation of the drawing]

FIG. 1 is a block configuration diagram showing the basic configuration of the present invention. FIG. 2 is a basic flow chart of the present invention. FIG. 3 is a block diagram of the first embodiment of the present invention. FIG. 4 is a flowchart showing the operation of the first embodiment. FIGS. 5(a) to 5(c) are diagrams showing examples of display images in the first and second embodiments of the present invention. FIGS. 6(a) to 6(C) are diagrams showing changes in label values in the first embodiment. FIG. 7 is a diagram showing label change processing in the first embodiment. FIG. 8 is a block diagram of a second embodiment of the present invention. FIG. 9 is a flowchart showing the operation of the second embodiment of the present invention. FIGS. 10(a) to 10(C) are diagrams showing changes in label values according to the present invention. FIG. 11 is a diagram showing label change processing in the second embodiment of the present invention. FIGS. 12(a) to 12(e) are explanatory diagrams of designated areas in modified examples of the present invention. DESCRIPTION OF SYMBOLS 11... Control part, I2... Storage part, 13... Labeling means, 14... Interested pixel selection means, 15...
Label changing means, 16... Completion determining means, 17...
Pixel detection means.

Claims (1)

[Scope of Claims] Labeling means for labeling an area including at least the boundary of an object; pixel-of-interest selection means for selecting a pixel from a specified area including the boundary of the object; and a label for the selected pixel of interest. a label value changing means for changing the label value of a pixel of interest based on the value and feature, and the label value and feature of a group of pixels in the vicinity of the pixel of interest; An image processing device characterized by having a completion determination means for terminating the change, and a pixel detection means for detecting pixels having a label that satisfies a specific condition from a label image as pixels belonging to a boundary, inside, or outside of the object. .