KR100217148B1 - A method of reducing mosquito noise generated during decoding processing of video data and a video data decoding device using the method - Google Patents

Abstract

An apparatus for decoding decoded image data, which generates decoded image data by applying inverse orthogonal transformation to each block in which coded image data is divided, provides a quantization information detection unit (5, 9) for detecting a quantization step size used for a given block. ), A frequency characteristic modifying unit (7, 7A) for modifying the high frequency component of the decoded image data in a given block based on the block quantization step size information, and a high frequency component having a frequency higher than a predetermined frequency.

Description

A method for reducing mosquito noise generated during decoding of image data and an apparatus for decoding image data using the method

1 is a block diagram of an image data decoding apparatus according to a first embodiment of the present invention.

2 is a block diagram of the frequency characteristic correction unit of FIG.

3 is a block diagram of another embodiment of a frequency characteristic correction unit.

4 is a diagram of an example characteristic curve showing the relationship between a coring level coefficient and a quantization scale factor.

5 is a block diagram of a video data decoding apparatus according to a second embodiment of the present invention.

6 is a block diagram of a video data decoding apparatus according to a third embodiment of the present invention.

FIG. 7 is an exemplary view for explaining a method of determining a leveling level for a pixel array in the control signal generating unit of FIG.

8 is a block diagram of a video data decoding apparatus according to a fourth embodiment of the present invention.

9 is a block diagram of a video data decoding apparatus according to a fifth embodiment of the present invention.

10 is a block diagram of another configuration of the frequency characteristic correction unit of FIG.

11 is a block diagram of another configuration of the frequency characteristic correction unit of FIG.

12 shows an embodiment for calculating an activity value.

FIG. 13 is a diagram illustrating an embodiment of a relationship between a multiplication factor and a value G. FIG.

14 is a block diagram of a video data decoding apparatus according to a sixth embodiment of the present invention.

* Explanation of symbols for main parts of the drawings

7: frequency characteristic correction unit 6: control signal generator

22: Subtractor 23: Multiplier

24: adder 25: vertical LPF

The present invention relates to a method for reducing noise and an image decoding apparatus, in particular, to a method for reducing mosquito noise occurring during a process of decoding encoded image data, and an apparatus for decoding encoded image data based on the method. will be.

When transmitting, reproducing, and recording various signals such as video signals, audio signals, and the like as digital signals, techniques for compressing and decompressing information are generally used to reduce the amount of information, that is, the number of bits. For example, if linear quantization (equal quantization) representing each sample value to a selected value from uniformly divided signal levels is used to digitize audio signals, video signals, etc. without any compression technique, recording / playback or transmission The amount of information that needs to be increased will probably be large. Thus, in the field of recording / reproducing information, communication, and broadcasting, human visual and audio perceptual characteristics have been used in compression techniques. For example, human perception is sensitive to changes in signal level when the signal has small fluctuations, but not sensitive to changes in signal level when the signal has strong fluctuations. These characteristics can be used to reduce the amount of information for each sample value. In addition, a number of techniques for compressing information have been used to effect the development of compression techniques.

For example, the amount of information included in a one-hour moving picture having a video quality similar to that reproduced by a VHS-type VTR device is about 109 gigabits. In addition, more than or equal to 360 gigabit is included in the one-hour moving picture with a picture quality comparable to the received picture of an NTSC color television set. Efforts to develop highly efficient compression techniques relate to applied research aimed at recording / reproducing or transmitting large amounts of information by current transmission lines or recording media.

The most efficient compression method proposed as a practical method applicable to image information is combined with three different compression techniques to reduce the amount of information. The first technique reduces the amount of information by using the correlation in the image frame taking advantage of the high correlation between adjacent pixels of the original image (compression using spatial correlation). The second technique reduces the amount of information by using correlations between temporally arranged image frames (compression using time correlations). The third technique reduces the amount of information by using different probabilities that each code can occur. As a technique (first technique) for compressing image information by using correlations in an image frame, various techniques have been proposed. Recently, orthogonal transforms such as the Karhunen-Loeve (K-L) transform, the Discrete Cosine transform (DCT), the Discrete Fourier Transform, and the Walsh-Hadam Mad transform are often used.

For example, a high-efficiency encoding method for image information proposed by MPEG (Moving Picture Experts Group) established under 150 (International Standards Organization) uses two-dimensional DCT. These highly efficient coding methods (MPEGI, MPEG2) use intra-frame coding and inter-frame coding to realize efficient coding of moving picture information while using motion compensated prediction and inter-frame prediction. coding). Orthogonal transformation is generally applied to blocks generated by dividing an image into unit blocks having a predetermined block size (M x N). In MPEG1 and MPEG2, one block having an 8x8 pixel block size is defined as one unit block.

The M × N orthogonal transform coefficients (eg, 64 DCT transform coefficients in MPEG1 and MPEG2) obtained by orthogonal transforming the unit blocks are quantized by using a block quantization step size (interval for quantization). The block quantization step size is defined for a region of a predetermined size including at least one unit block. In MPEG1 and MPEG2, for example, a region of a predetermined size is called a macro block, which is made up of blocks of 16x16 pixels for the luminance signal Y and 8x8 for each color signal Cr and Cb. It consists of a block of pixels. In detail, the block quantization step size is expressed as [the quantization characteristic of the macro block (quantization scale of the macro block) QS × quantization matrix (8 × 8)]. Here, the quantization characteristic of the macro block changes for each macro block.

The orthogonal transform coefficients (e.g., DCT coefficients) quantized based on the block-quantization step size are separated into direct current components (DC components) and alternating current components (AC components). The DC component of the orthogonal transform coefficient is affected by differential coding, and the AC component of the orthogonal transform coefficient is affected by entropy coding after zigzag scanning. Here, entropy coding is an information compression technique using a variable length coding scheme that uses various probabilities of generating each code as in Huffman coding. The converted and encoded video data is transmitted as a bit stream (serial bit). The decoding operation on the transformed and encoded image data is performed in the inverse manner of the above-described encoding operation to generate an output image. However, when the quantization process is included in the entire encoding process, quantization noise is caused in the output image due to the inevitable quantization error. Thus, quantization noise will degrade image quality when it contains more information than the capacity of the complexity rate of the image processed in the encoding process.

In general, the quantization error in the low frequency component causes block distortion in the output image, and there is no correlation between each block of the output image by this block distortion. In addition, quantization errors in high frequency components generate mosquito noise around the edges, which is a distortion in which ringing appears in the output image.

Quantization errors appearing in the output image are particularly severe where the image level is generally flat. When a small amount of quantization noise is added at the time when the change in the video signal level has frequency components from low frequency to high frequency, the noise is difficult to be detected because of visual perceptual characteristics. However, when a small amount of noise having a high frequency component is added at the time when the change of the video signal has only a low frequency component, the noise can be easily detected. Of course, when a large amount of noise is added, the noise is detected as a coding degradation regardless of the frequency component of the noise.

As a measurement technique for calculating quantization noise, there is a known technique called a coring technique. This technique assumes that a portion of the decoded image having a small high frequency component is composed mostly of image signal noise and quantization noise, and that a portion of the image having a signal level lower than a predetermined level has a zero signal level. This ringing technique has been used as a method of removing small noise in analog signals (see, eg, Broadcasting Technology, pages 141-147, February 1991).

In general, the block quantization step size has a larger value for high frequency components than for low frequency components, reflecting visual perceptual characteristics. Therefore, quantization error tends to appear in high frequency components. In view of this, the above-described coring technique is applied only to the high frequency component to convert the signal level lower than the predetermined level to zero level. At this time, the signal processed by the ringing technique may be added to the signal of the low frequency component. However, when the ringing level is set in accordance with the magnitude of the quantization noise generated by the large quantization step size, the output image tends to appear in a flat form without fine texture. On the other hand, if a coring level corresponding to the amount of quantization noise generated by the small quantization step size is used, the quantization noise having a large magnitude cannot be removed. In other words, since the ringing level is fixed at a predetermined level, the quantization noise is not reduced efficiently.

Accordingly, there is a need for a method for efficiently reducing mosquito noise generated during decoding of encoded image data and an apparatus for decoding encoded image data based on the method.

Accordingly, it is an object of the present invention to provide a method and apparatus that can satisfy the aforementioned needs.

Another object of the present invention is to provide a method for efficiently reducing mosquito noise generated during a decoding process of encoded image data, and an apparatus for decoding encoded image data based on the method.

In order to achieve the above object of the present invention, a decoding apparatus for generating an image data decoded by inverse orthogonal transformation of each block of encoded image data divided into a plurality of blocks includes a block indicating a quantization step size used for a given block. A quantization information detection unit for detecting quantization step size information, and a frequency characteristic correction unit for removing high frequency components (mosquito noise) of the decoded image data for the given block based on the block quantization step size information, The high frequency component has a frequency higher than a predetermined frequency.

In addition, the same object is achieved by the decoding method according to the present invention which generates decoded image data by inverse orthogonal transformation of each block of encoded image data divided into a plurality of blocks. The method includes detecting block quantization step size information indicative of the quantization step size used for a given block, and removing high frequency components of the decoded image data for the given block based on the block quantization step size information. And the high frequency component has a frequency higher than a predetermined frequency.

In the above-described method and apparatus, block quantization step size information indicative of the quantization step size used for a given block is used to remove high frequency components of the decoded image data. That is, mosquito noise having a magnitude corresponding to the quantization step size can be reduced by removing high frequency components in accordance with the quantization step size. Therefore, the mosquito noise is efficiently reduced without adversely affecting the fine shape of the decoded image data.

Other objects and further features of the present invention will be described in detail below with reference to the drawings.

1 is a block diagram of an image data decoding apparatus according to a first embodiment of the present invention. In FIG. 1, an input node receiving a bit stream (a series of bits) to be decoded is indicated by reference numeral 1. In addition, the part enclosed by the dashed-dotted line 3 is provided to an integrated circuit. The portion enclosed by the dashed line 3 includes a buffer memory 8, a variable length decoding unit 9, an inverse quantization unit 10, an inverse orthogonal transform unit 11, an adder 12, a motion compensation unit 13 and And an image memory 14. Integrated circuits for this part are commercially available.

The bit stream provided to the input node 1 uses the three different compression techniques described above, that is, a compression technique using spatial correlation (compression technique using spatial correlation), and a correlation between temporally arranged image frames. It is image data encoded through a highly efficient encoding method such as MPEG1 or MPEG2, which combines a compression technique (compression technique using time correlation) and a compression technique using various probabilities in which each code can be generated. In the description described below, it is assumed that video data to be decoded is generated by MPEG1 or MPEG2.

Highly efficient encoding of moving picture information over MPEG1 or MPEG2 is performed by a combination of intra-frame coding and inter-frame coding through two-dimensional discrete cosine transform (two-dimensional DCT), and also motion compensation prediction and inter-frame prediction. Use The video signal for each image processed by highly efficient encoding is divided into unit blocks having an 8 × 8 pixel block size (8 pixels in the horizontal direction and 8 lines in the vertical direction), and DCT is applied to each unit block. do. The 64 DCT transform coefficients obtained for each unit block are then quantized by using the block quantization step size. In MPEG1 and MPEG2, a block quantization threshold is defined for each macro block, each macro block being a predetermined size region containing at least one unit block, and each of 16 × 16 pixel blocks and each of the luminance signals Y It consists of blocks of 8x8 pixels for the color signals Cr and Cb. In detail, the block quantization step size is expressed as [the quantization characteristic of the macro block (quantization scale of the macro block) QS × quantization matrix (8 × 8)]. Here, the quantization characteristic of the macroblock (quantization scale of the macroblock) QS is a scale factor that varies from macroblock to macroblock.

The DCT coefficients obtained through the quantization process of dividing the DCT transform coefficients by the block quantization step size are separated into a DC component (DC component) and an AC component (AC component). The DC components of the DCT coefficients are affected by differential coding, and the AC components of DCT coefficients are affected by entropy coding (variable length boring schemes that use different probabilities for each code generation, as in Huffman coding). Then, a bit stream in which information necessary for the decoding process is added to the converted and encoded image data is generated. Here, the information necessary for the decoding process includes information about a block quantization step size (eg, quantization scale QS of a macro block), and information about a motion vector and a prediction mode. In the video data decoding apparatus of FIG. 1, the bit stream supplied to the input node 1 is stored in a buffer memory 8 made from a first-in first-out (F1F0) memory.

The variable length decoding unit 9 receives the bit stream read into the buffer memory 8 and stores the encoded image data and the supplementary information necessary for decoding the transformed and encoded image data through entropy encoding (variable length encoding). Decode (The auxiliary information includes block quantization step size information (quantization characteristic QS of the macro block), and information about a motion vector and a prediction mode.

At this time, the image data and the block quantization step size information (quantization scale QS of the macro block) decoded by the variable length decoding unit 9 are supplied to the inverse quantization unit 10. In addition, information regarding the motion vector and the prediction mode is supplied to the motion compensation unit 13.

The inverse quantization unit 10 receiving the image data and the block quantization step size information QS obtains the DCT transform coefficients through the inverse quantization process and provides the coefficients to the inverse orthogonal transform (inverse DCT) unit 11. The inverse orthogonal transform (inverse DCT) unit 11 applies a two-dimensional inverse DCT to each unit block in order to convert the image data in the frequency domain into the image data in the time domain. The result of the image data in the time domain is supplied to the adder 12. In the adder 12, the image data in the time domain can be added to the motion compensation data obtained by the motion compensation unit 13 according to the encoding form representing one of intra-frame encoding and inter-frame encoding. The image data output from the adder 12 is stored in the image memory 14.

In the image data decoding apparatus of FIG. 1, the image data provided from the image memory 14 is output to the output node 2 via the frequency characteristic correction unit 7. The frequency characteristic correction unit 7 is a coring circuit for performing the coring process.

The frequency characteristic correction unit 7 is controlled by a ringing level control signal provided from the control signal generating unit 6 to change the ringing level applied to the image data. 2 is a block diagram of the frequency characteristic correcting unit 7. 3 is a block diagram of another example of the frequency characteristic correcting unit 7. One of the frequency characteristic correction units 7 shown in FIGS. 2 and 3 can be used for the image data decoding unit of FIG.

In FIG. 2, the portion surrounded by the dotted line 7h is used as a coring circuit for applying the coring process to the horizontal direction of the image. In addition, the portion surrounded by the dotted line 7v is used as a coring circuit for applying the coring process to the vertical direction of the image. The dotted line used as the coring circuit for the horizontal direction and the dotted line used as the coring circuit for the vertical direction. The part enclosed by 7v is connected in series. Thus, both parts form a coring circuit for applying the two-dimensional coring process.

In FIG. 2, the portion enclosed by the dotted line 7h includes a horizontal LPF 21, a subtractor 22, a multiplier 23 and an adder 24 having a predetermined cutoff frequency. Further, the portion surrounded by the dotted line 7v includes a vertical LPF 25, a subtractor 26, a multiplier 27 and an adder 28 having a predetermined cutoff frequency.

In FIG. 3, the frequency characteristic correction unit 7 includes a two-dimensional LPF 30, a subtractor 31, a multiplier 32, and an adder 33, and is used as a coring circuit for applying a two-dimensional coring process. do. When the filter is applied in two dimensions, the amount of calculation for addition and multiplication operations increases. However, there is an advantage in that the frequency band extending in the diagonal direction can be processed.

In FIG. 2, the multiplier 23 in the portion surrounded by the dotted line 7h and the multiplier 27 in the portion surrounded by the dotted line 7v are provided with the coring level control signal from the control signal generating unit 6. In FIG. 3, the multiplier 32 is provided with a coring level control signal from the control signal generation unit 6.

In the multiplier of the frequency characteristic correcting unit 7, signal components in a frequency band higher than a predetermined cutoff frequency of the LPF are multiplied by zero when the magnitude of the signal component is lower than the coring level control signal (coring level coefficient signal). Thus, the coring process is performed to reduce undesirable signal components in the highest frequency band.

4 is a diagram of an example of a characteristic curve showing the relationship between the coring level coefficients and the block quantization step size information QS.

Coring level coefficients shown in the vertical coordinate axis of FIG. 4 are from 0 to 10, and frequency characteristic correction for performing signal processing to suppress high frequency components according to block quantization step size information (e.g., quantization scale QS of a macroblock). Used by the unit 7. Here, the block quantization step size information is defined for each macro block, that is, for each predetermined size region including at least one unit block (DCT block) to which DCT is applied. In the embodiment of FIG. 4, the coring level coefficient increases as the value of the block quantization step size information increases and saturates at any level.

The control signal generating unit 6 used in the video data decoding apparatus of FIG. 1 is a look-up table in which block quantization step size information (for example, the quantization scale QS of the macro block) defined by caching in a macro block is provided as an address. (ROM tables). The look-up table (ROM table) outputs a coring level coefficient signal (coring level control signal) corresponding to the coring level coefficient according to the address provided to it.

The frequency characteristic correction unit 7 which performs the coring process according to the coring level control signal provided from the control signal generating unit 6 uses a large coring level in the image region in which a large block quantization step size is used. Thus, the mosquito noise can be easily removed while maintaining the sharp edges of the image.

That is, it is possible to take advantage of the characteristics that an image region using a large quantization step size tends to generate a large amount of mosquito noise, and an image region using a small quantization step size tends to generate a small size of mosquito noise. . Therefore, the frequency characteristic correction unit 7 performs an adaptive coring process that is enlarged in an image region using a block quantization step size having a large coring level. As a result, it is possible to prevent the fine shape required for the image from disappearing during the mosquito noise removing process.

In the video data decoding apparatus of FIG. 1 using the adaptive coding process of the present invention, the bit stream (a series of bits) provided to the input node 1 has a buffer memory 8 and a variable length decoding unit 9 as described above. The decoded image data is decoded by the inverse quantization unit 10, the inverse orthogonal transform unit 11, the adder 12, and the motion compensation unit 13. The decoded image data is stored in the image memory 14 and then output from the image memory 14.

The image data read out from the image memory 14 is provided to the frequency characteristic correction unit 7 in which the quantization noise is reduced by the adaptive coring process. Therefore, the image data having the suppressed quantization noise is output at the output node 2 of the image data decoding apparatus provided to the adaptive coding apparatus using the adaptive coding process.

In the image data decoding apparatus of FIG. 1 with an adaptive coding device, the bit stream provided to the input node 1 is stored in the buffer memory 8 of the decoder integrated circuit surrounded by the dashed line 3, and also the buffer memory 4 Is supplied. The bit stream read from the buffer memory 4 of the first-in-first-out memory is supplied to the block quantization step size information detection unit 5. For the block quantization step size information detection unit 5, a mechanism similar to the variable length decoding unit 9 can be used.

The block quantization step size information detection unit 5 detects the block quantization step size information (quantization scale QS of the macro block) and provides it to the control signal generation unit 6.

That is, in the video data decoding apparatus of FIG. 1, the bit stream provided to the input node 1 is provided to the block quantization step size information detection unit 5 through the buffer memory 4, and this block quantization step size information detection unit ( In 5), block quantization step size information is detected from a bit stream for a series of macro blocks to be provided to the control signal generating unit 6. Here, the buffer memory 4, the block quantization step size information detection unit 5, the control signal generation unit 6, and the frequency characteristic correction unit 7 are provided outside the decoder integrated circuit surrounded by the dashed line 3. However, these units may be included in the same integrated circuit.

5 is a block diagram of an image decoding apparatus according to a second embodiment of the present invention. In FIG. 5, the same components as those in FIG. 1 are referred to by the same reference numerals, and description thereof is omitted.

In the image data decoding apparatus of FIG. 5 equipped with an adaptive coding device, all units are provided in a decoder integrated circuit surrounded by a dashed line 15.

Further, the block quantization step size information detected by the variable length decoding unit 9 from the bit stream for the series of macro blocks is provided to the control signal generating unit 6.

That is, in the image data decoding apparatus of FIG. 5 equipped with an adaptive coding apparatus, the block quantization step size information detection unit 5 and the buffer memory 4 of FIG. 1 provided outside the decoder integrated circuit surrounded by the dashed line 3 are provided. The operation of is performed by the variable length decoding unit 9 and the buffer memory 8 provided in the decoder integrated circuit surrounded by the dashed line 15. Also in FIG. 5, the decoder integrated circuit surrounded by the dashed line 15 includes a control signal generating unit 6 and a frequency characteristic correcting unit 7. Here, the control signal generation unit 6 is provided with block quantization step size information (quantization scale QS of the macro block) extracted from the bit stream for the series of macro blocks by the variable length decoding unit 9. At this time, the frequency characteristic correction unit 7 suppresses the high frequency content of the image data provided from the image memory 14 on the basis of the ringing level control signal supplied from the control signal generation unit 6.

As described above, in the first embodiment and the second embodiment of the present invention, the ringing level control signal generated by the control signal generating unit 6 is provided to the frequency characteristic correcting unit 7. As shown in FIG. 4, the coring level control signal is associated with each macro block quantization step size information included in the bit stream as auxiliary information. At this time, the frequency characteristic correction unit 7 suppresses the high frequency component having a level lower than the coring level determined according to the block quantization step size information.

6 is a block diagram of an image data decoding apparatus according to a third embodiment of the present invention. In FIG. 6, the same components as those of FIG. 1 are referred to by the same reference numerals, and description thereof is omitted.

In the video data decoding apparatus of FIG. 6 equipped with the adaptive coding apparatus of the present invention, the bit stream to be decoded is supplied to the buffer memory 8 of the decoder integrated circuit surrounded by the dashed line 16 through the input node 1. In addition, the bit stream supplied to the input node 1 is provided to the buffer memory 4. The bit stream read from the buffer memory 4 which is a first-in first-out memory is supplied to the block quantization step size information detection unit 5. The block quantization step size information detection unit 5 may use a mechanism similar to the variable length decoding unit 9. The block quantization step size information detection unit 5 detects the necessary information from the bit stream for each macro block and provides it to the control signal generation unit 17.

The control signal generating unit 17 used in the video data decoding apparatus of FIG. 6 may include a look-up table (ROM table). The look-up table receives, as its address, the block quantization step size information defined for each macro block from the block quantization step size information detection unit 5 and the position of the pixel currently processed from the image memory 14. Here, as described above, the macro block includes at least one unit block to which the DCT is applied. In the control signal generation unit 17, the look-up table (ROM table) is a coring level coefficient signal (coring) corresponding to the coring level coefficient according to the address provided with (block quantization step size information and the position of the currently processed pixel). Level control signal).

The reason why the position of the currently processed pixel is provided to the control signal generating unit 17 together with the block quantization step size information is that an increase in the ringing level near the boundary of the unit block can effectively reduce the block distortion. In other words, by changing the coring level according to the block quantization step size information, not only can the mosquito noise be efficiently reduced, but also the block distortion can be efficiently reduced, which is the pixel near the boundary of the unit block according to the position of the currently processed pixel. This is done by increasing the level of coring for.

FIG. 7 is a diagram for explaining a method of determining a leveling level for the pixel array in the control signal generating unit 17. As shown in FIG. Here, the control signal generation unit 17 generates a coring level control signal to efficiently reduce block distortion by setting the coring level to a large level for the pixels near the boundary of the unit block. In FIG. 7, 11, 12, 13 and 14 represent different time points, each of which is delayed by one clock interval. In a given one-line pixel array of images, the predetermined coring level is set for pixels p-3 to p + 6 located near the unit block boundary.

As described above, in the third embodiment of the present invention, the control signal generation unit 17 includes the block quantization step size information detection unit 5 from the block quantization step size information and the pixels currently processed from the image memory 14. Location (address) is provided. At that time, the ROM table of the control signal generating unit 17 is a coring level coefficient signal (coring level control signal) corresponding to the coring level coefficient according to the address provided with (block quantum step size information and the position of the currently processed pixel). Outputs Therefore, the frequency characteristic correction unit 7 can increase the coring reel for pixels located near the boundary of the unit block in order to effectively reduce not only mosquito noise but also block distortion.

8 is a block diagram of an image data decoding apparatus according to a fourth embodiment of the present invention. In FIG. 8, the same components as those of FIG. 6 are referred to by the same reference numerals, and description thereof is omitted.

The video data decoding apparatus of FIG. 8 equipped with the adaptive coding apparatus differs from the video data decoding apparatus of FIG. 6 as follows. The block quantization step size information (for example, the quantization scale QS of the macro block) provided to the control signal generation unit 17 is detected from the bit stream by the variable length decoding unit 9. In addition, the control signal generation unit 17 and the frequency characteristic correction unit 7 are provided in a decoder integrated circuit surrounded by a dashed line 18.

As can be seen from the foregoing description, according to the first to fourth embodiments of the present invention, the block quantization step size information is increased such that the level of coring used in the coring process is increased for an image region using a large block quantization step size. Can change accordingly. Therefore, the mosquito noise can be easily reduced while maintaining the image edge. In addition, the fact that large magnitude mosquito noise appears in the compressed image region using the coarse quantization step and small magnitude mosquito noise appears in the region using the fine quantization step can be used. That is, the frequency characteristic correcting unit 7 to which the adaptive coding method is applied is coring in the region of a large block quantization step size according to the coring level control signal provided from the control signal generating unit so that the fine shape of the image is not damaged by mosquito noise removal. Increase the level. Moreover, the coring level can be increased for pixels near the boundary of the unit block so that block distortion is effectively reduced with mosquito noise.

9 is a block diagram of an image data decoding apparatus according to a fifth embodiment of the present invention. In FIG. 9, the same components as those in FIG. 1 are referred to by the same reference numerals, and description thereof is omitted.

In the image data decoding apparatus of FIG. 9, the first activity detection unit 40 and the second activity detection unit 41 are provided to obtain an activity value (a measure of image complexity) of the image data stored in the image memory 14. Is different from the image data decoding apparatus of FIG. The activity values obtained by the first activity detection unit 40 and the second activity detection unit 41 are supplied to the control signal generation unit 6A. The control signal generation unit 6A generates a control signal based on the active value and provides it to the frequency characteristic correction unit 7A.

10 is a block diagram of the frequency characteristic correcting unit 7A, and FIG. 11 is a block diagram of another configuration of the frequency characteristic correcting unit 7A. The configuration of FIGS. 10 and 11 used in the fifth embodiment is different from the configuration of FIGS. 2 and 3 used only in the first to fourth embodiments in the control operation of the multiplier. That is, the multipliers of FIGS. 2 and 3 multiply zero by the magnitude of the high frequency component when the magnitude of the high frequency component is less than the adaptive coling level. On the other hand, multipliers 23A and 27A in FIG. 10 and multiplier 32A in FIG. 11 multiply the magnitudes of the high frequency components by the control signal provided thereto.

Multipliers 23A and 27A (or multiplier 32A in FIG. 11) of FIG. 10 receive a control signal from control signal generating unit 6A, and the control signal is a multiplier signal having a range of 0 to 1,25. In the output image data of the frequency characteristic correction unit 7A, signal components having frequencies higher than the predetermined cutoff frequencies of the LPFs 21 and 25 (30) are multiplied by a multiplier having a range of 0 to 1.25.

When the control signal supplied from the control signal generating unit 6A to the frequency characteristic correcting unit 7A is in the range of 0 to 1.0, the image data appearing at the output node 2 is at a frequency higher than the predetermined cutoff frequency of the LPF. It has a signal component that is attenuated by a factor ranging from 0 to 1.0.

When the control signal supplied from the control signal generating unit 6A to the frequency characteristic correcting unit 7A is in the range of 1.0 to 1.25, the image data appearing at the output node 2 is at a frequency higher than the predetermined cutoff frequency of the LPF. It has a signal component that is raised by a factor ranging from 1.0 to 1.25. Thus, the outline in the image is improved.

The control signal supplied to the frequency characteristic correcting unit 7A is generated by the control signal generating unit 6A in the following manner. In FIG. 9, the control signal generation unit 6A receives the block quantization step size information (e.g., the quantization scale QS of the macro block) from the block quantization step size information detection unit 5, and the first activity detection unit ( 40) and an activity value from the second activity detection unit 41. The first activity detection unit 40 detects a first activity value A1 for a region having a first size in the decoded image region. The second activity detection unit 41 detects a second activity value A2 for the region having a second size smaller than the first size in the decoded image region.

The first activity detecting unit 40 and the second activity detecting unit 41 have a function of detecting activity (a measure of image complexity) for a predetermined size region in the image. The activity value of the image detected by the first activity detection unit 40 and the second activity detection unit 41 may be any one of various conventional measures used as the activity value of the image.

That is, the first activity detecting unit 40 and the second activity detecting unit 41 may detect the sum of the differences between adjacent pixels as the active value as shown in FIG. 12.

Alternatively, the first activity detection unit 40 and the second activity detection unit 41 may add the average value of the difference between the pixel value and the average pixel value in the block and the absolute value of the pixel value after applying a predetermined filtering process. The deviation obtained can be detected by calculating.

The region having the first size in the decoded image region in which the first activity detection unit 40 detects the first activity value Al may be a unit block (DCT block). In the decoded image area in which the second activity detection unit 41 detects the second activity value A2, the area having the second size smaller than the first size is the filter of the FIR filter used in the frequency characteristic correction unit 7A. It may have a plurality of pixels corresponding to the length. For example, if the FIR filter is a one-dimensional filter having three taps or five taps as the filter length, the region having the second size may have three pixels or five pixels, respectively. In addition, if the FIR filter is a two-dimensional filter having three taps or five taps as the filter length, the region having the second size may have 3 × 3 pixels or 5 × 5 pixels, respectively.

Control signal generation unit 6A based on the quantized block level size information (for example, the quantization scale QS of the macro block), the first active value A1 and the second active value A2 defined for each macro block. ) Is calculated as follows:

G = {K1- (QS × (A1 / A2) +1)} / K2 (1)

Here, K1 and K2 are constants.

Then, the control signal generating unit 6A uses the value G as an address of the ROM table to generate a multiplication factor signal having a range of 0 to 1.25, for example.

13 is a diagram showing an example of the relationship between the multiplication factor and the value G. FIG.

As shown in Fig. 13, the multiplication factor is saturated at the upper limit when the value G exceeds the predetermined value Th. In FIG. 13, the upper value appears to have a value of 1.25. However, the upper limit does not need to be fixed at 1.25. As mentioned above, the multiplication factor signal is supplied to the multipliers 23A, 27A (32A) of the frequency characteristic correcting unit 7A.

In equation (1), the value G decreases as the block quantization step size information defined for each macro block (e.g., the quantization scale QS of the macro block) increases or increases (A1 / A2). Here, the block quantization step size information increases as the image becomes more complicated.

The value of (A1 / A2) indicates whether the image data near a given pixel has a similar complexity as the image data around it. That is, the value of (A1 / A2) indicates whether a given pixel is a pixel of a flat portion of an image or a pixel of an edge portion of an image.

For example, the value of (A1 / A2) decreases at the edge portion so that the value G becomes larger. In contrast, the flat portion of the image has a large value (A1 / A2) so that the value G becomes small.

In the following, the operation of the frequency characteristic correcting unit 7A and the result of the operation will be examined so that the fifth embodiment can be easily understood.

When the control signal supplied to the multipliers 23A and 27A (32A) of the frequency characteristic correcting unit 7A matches a multiplier equal to 0, the image data appearing at the output node 2 from the frequency characteristic correcting unit 7A. Has only a frequency component corresponding to the frequency bands of the LPFs 21 and 25 (30). That is, the image data does not have a signal component having a frequency higher than the predetermined cutoff frequency of the LPFs 21 and 25 (30).

When the control signal supplied to the multipliers 23A and 27A (32A) of the frequency characteristic correcting unit 7A matches the multiplier equal to 0.5, the image data appearing at the output node 2 from the frequency characteristic correcting unit 7A. Has a frequency component corresponding to the frequency bands of the LPFs 21 and 25 (30). Moreover, the image data has a signal component having half the magnitude of the original signal component having a frequency higher than the predetermined cutoff frequency of the LPFs 21 and 25 (30).

Image data appearing at the output node 2 from the frequency characteristic correction unit 7A when the control signal supplied to the multipliers 23A, 27A (32A) of the frequency characteristic correction unit 7A matches the multiplier equal to 1.0. Is the same as the image data supplied to the frequency characteristic correction unit 7A.

When the control signal supplied to the multipliers 23A and 27A (32A) of the frequency characteristic correcting unit 7A matches the multiplier equal to 1.25, the image data appearing at the output node 2 from the frequency characteristic correcting unit 7A. Has a frequency component corresponding to that of the LPFs 21 and 25 (30). The image data has a signal component increased by 1.25 times the original signal component having a frequency higher than the predetermined cutoff frequency of the LPFs 21 and 25 (30). Thus, the resulting image has an improved edge.

14 is a block diagram of an image data decoding apparatus according to a sixth embodiment of the present invention. In FIG. 14, the same components as those of FIG. 9 are referred to by the same reference numerals, and description thereof is omitted.

In the image data decoding apparatus of FIG. 14, all units are provided in a decoder integrated circuit surrounded by a dashed line 17. FIG. Further, the block quantization step size information detected by the variable length decoding unit 9 from the bit stream for the series of macro blocks is provided to the control signal generating unit 6A.

That is, in the image data decoding apparatus of FIG. 14, the operation of the buffer memory 4 and the block quantization step size information detection unit 5 of FIG. 9 provided outside the decoder integrated circuit indicated by the dashed line 3 is performed by the dashed line. It is performed by the buffer memory 8 and the variable length decoding unit 9 provided in the decoder integrated circuit enclosed by (17). Also in FIG. 14, the decoder integrated circuit surrounded by the dashed line 17 includes a control signal generating unit 6A and a frequency characteristic correcting unit 7A.

Here, the control signal generating unit 6A is provided with block quantization step size information (e.g., macro bleed quantization scale QS) extracted by the variable length decoding unit 9 from the bit stream for the series of macro blocks. . At this time, the frequency characteristic correction unit 7A corrects the high frequency component of the image data provided from the image memory 14 based on the control signal supplied from the control signal generation unit 6A.

As can be seen from the above description, according to the fifth and sixth embodiments according to the present invention, the first active detection unit receives the decoded image data from the image memory and has an area having a first size in the decoded image. Detect a first activity value for and provide a first activity value to the control signal generating unit. The second activity detection unit also detects a second activity value for the area having a second size less than the first size and provides a second activity value to the control signal generating unit.

Based on the first and second active values and the block quantization step size information, the control signal generating unit generates a control signal used as a multiplication factor by the multiplier of the frequency characteristic correcting unit. The frequency characteristic correction unit attenuates signal components having a high frequency in an image region having a large block quantization step size and a small active value. In contrast, the frequency characteristic correction unit has a high frequency in an image region having a small block quantization step size and a large active value. Increase signal component Therefore, the mosquito noise generated during the decoding of the image data is effectively reduced, and the edge portion of the image is improved to increase the image quality.

In addition, the present invention is not limited to these embodiments, and other modifications may be made without departing from the scope of the present invention.

Claims (24)

A decoding apparatus for generating decoded image data by inverse orthogonal transformation of each block of encoded image data divided into a plurality of blocks, the block quantization detecting block quantization step size information indicating a quantization step size used for a given block. Step size information detection units 5 and 9; And frequency characteristic correction units (7, 7A) for removing high frequency components of the decoded image data for the given block based on the block quantization step size information, wherein the high frequency components have frequencies higher than a predetermined frequency. Decoding apparatus, characterized in that. 2. The control circuit according to claim 1, further comprising a control signal generation unit (6) for generating a control signal based on the block quantization step size information, wherein the frequency characteristic correction unit (7) is based on the control signal. A decoding apparatus for removing the high frequency component of the decoded image data with respect to 3. The frequency characteristic correcting unit (7) according to claim 2, further comprising: low frequency separating means (21, 25, 30) for separating low frequency components of the decoded video data; High frequency separation means (22, 26, 31) for separating the high frequency components; Removing means (23, 27, 32) for removing the high frequency component when the magnitude of the high frequency component is smaller than the control signal value; Coupling means (24, 28, 33) for coupling said low frequency component with the output of said removal means (23, 27, 32), said low frequency component having a frequency lower than said predetermined frequency; Decryption device. The variable length decoding means (9) according to claim 2, wherein said block quantization step size information detecting units (5, 9) decode the encoded image data through a variable length decoding scheme and generate the block quantization step size information. Decryption apparatus comprising a. The frequency characteristic correction unit (7) according to claim 2, characterized in that the frequency characteristic correction unit (7) is adapted to remove the high frequency component of the decoded image data based on the block quantization step size information and the address of the currently processed pixel of the decoded image data. Decoding apparatus comprising a. 6. The apparatus according to claim 5, further comprising a control signal generating unit (17) for generating a control signal based on the block quantization step size information and the address of the currently processed pixel. A decoding device for removing the high frequency component of the decoded image data for the given block based on a control signal; 5. The frequency characteristic correcting unit (7) according to claim 4, further comprising: low frequency separating means (21, 25, 30) for separating low frequency components of said decoded video data; High frequency separating means (22, 26, 31) for separating the high frequency means; Removing means (23, 27, 32) for removing the high frequency component when the magnitude of the high frequency component is smaller than the value of the control signal; Coupling means (24, 28, 33) for coupling said low frequency component and the output of said removal means (23, 27, 32), said low frequency component having a frequency lower than said predetermined frequency; Decryption device. 6. The variable length decoding means (9) according to claim 5, wherein the block quantization step size information detection units (5, 9) decode the encoded video data through a variable length decoding method and generate the block quantization step size information (9). Decoding apparatus comprising a). A decoding apparatus for generating decoded image data by inverse orthogonal transformation of each block of encoded image data divided into a plurality of blocks, the block quantization detecting block quantization step size information indicating a quantization step size used for a given block. Step size information detection units 5 and 9; A frequency characteristic correction unit (7, 7A) for modifying the high frequency components of the decoded image data for the given block based on the block quantization step size information, wherein the high frequency components have frequencies higher than a predetermined frequency. ; And a first activity value detection unit 40 for detecting a first activity value for a first image area having a first size in the decoded image data, wherein the first activity value is an image in the first image area. A measure of complexity; A second activity value detection unit 41 for detecting a second activity value for a second image area having a second size smaller than a first size in the decoded image data, wherein the second activity value is Is a measure of the image complexity of a second image region: the frequency characteristic correction unit 7A is configured to determine the high frequency of the decoded image data based on the block quantization step size information, the first activity value, and the second activity value. A decoding device, characterized in that the component is corrected. 10. The apparatus according to claim 9, further comprising a control signal generating unit 6A for generating a control signal based on the block quantization step size information, the first active value, and the second active value. And (7A) operates according to the control signal to attenuate the high frequency component in an image region and to increase the high frequency component in another image region. 11. The frequency characteristic correcting unit (7A) according to claim 10, further comprising: low frequency separating means (21, 25, 30) for separating low frequency components of the decoded video data; High frequency separation means (22, 26, 31) for separating the high frequency components; Gain switching means (23A, 27A, 32A) for switching the magnitude of the high frequency component based on the control signal; Coupling means (24, 28, 33) for coupling said low frequency component and the output of said switching means (23A, 27A, 32A), said low frequency component having a frequency lower than said predetermined frequency; Decryption device. 10. The variable length decoding means (9) according to claim 9, wherein the block quantization step size information detecting means (5, 9) decodes the encoded video data through a variable length decoding method and generates the block quantization step size information (9). Decoding apparatus comprising a). A decoding method for generating decoded image data by inverse orthogonal transformation of each block of encoded image data divided into a plurality of blocks, the method comprising: a) detecting block quantization step size information indicating a quantization step size for a given block; ; b) removing a high frequency component of the decoded image data for the given block based on the block quantization step size information, wherein the high frequency component has a frequency higher than a predetermined frequency. . The method of claim 13, wherein step b) comprises: b1) generating a control signal based on the block quantization step size information; b2) removing said high frequency component of said decoded image data for said given block based on said control signal. 15. The method of claim 14, wherein step b2) comprises: b2-1) separating low frequency components of the decoded image data; b2-2) separating the high frequency component; b2-3) removing the high frequency component when the magnitude of the high frequency component is smaller than the value of the control signal; b2-4) combining the low frequency component and the output of step b2-3), wherein the low frequency component has a frequency lower than the predetermined frequency. The decoding method according to claim 13, wherein the step a) includes decoding the encoded image data and generating the block quantization step size information through a variable length decoding method. The method of claim 13, wherein the step b) removes the high frequency component of the protected image data based on the block quantization step size information and an address of a pixel currently processed in the decoded image data. Decryption method. 18. The method of claim 17, wherein b) comprises: b1) generating a control signal based on the block quantization step size information and the address of the currently processed pixel; b2) removing the high frequency component of the decoded image data for the given block based on the control signal. 19. The method of claim 18, wherein b2) comprises: b2-1) separating low frequency components of the decoded image data; b2-2) separating the high frequency component; b2-3) removing the high frequency component when the magnitude of the high frequency component is smaller than the value of the control signal; b2-4) combining the low frequency component and the output of step b2-3), wherein the low frequency component has a frequency lower than the predetermined frequency. 18. The decoding method according to claim 17, wherein the step a) comprises decoding the encoded image data and generating the block quantization step size information through a variable length decoding scheme. A decoding method for generating decoded image data by inverse orthogonal transformation of each block of encoded image data divided into a plurality of blocks, the method comprising: a) detecting block quantization step size information indicating a quantization step size for a given block; b) modifying a high frequency component of the decoded image data for the given block based on the block quantization step size information, wherein the high frequency component has a frequency higher than a predetermined frequency; c) detecting a first activity value for a first image region having a first size in the decoded image data, wherein the first activity value is a measure for image complexity in the first image region; d) detecting a second activity value for a second image area having a second size smaller than a first size in the decoded image data, wherein the second activity value is an image of the second image area. A measure of complexity: wherein step b) modifies the high frequency components of the decoded image data based on the block quantization step size information, the first activity value and the second activity value. The method of claim 21, further comprising: e) generating a control signal based on the block quantization step size information, the first activity value, and the second activity value, wherein step b) is performed in any image region. And attenuating high frequency components and amplifying the high frequency components in different image regions according to the control signal. 23. The method of claim 22, wherein step b) comprises: b2) b1) separating low frequency components of the decoded image data; b2) separating the high frequency component; b3) switching the magnitude of the high frequency component based on the control signal; b4) combining the low frequency component and the output of step b3), wherein the low frequency component has a frequency lower than the predetermined frequency. 22. The decoding method of claim 21, wherein the step a) includes decoding the encoded image data and generating the block quantization step size information through a variable length decoding method.