JP2006229411A - Image decoder and image decoding method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide technology of reducing flicker noise without causing an after-image at a low computation complexity by using intra prediction information in a stream so as to adaptively carry out control of post filtering. <P>SOLUTION: A filter control section 105 uses the intra prediction information to switch a temporal/spatial filter in accordance with the size of a block, and determines the characteristic of the temporal/spatial filter by using a temporal change in the intra prediction information. A filter processing section 106 applies the temporal/spatial filter to a decoded image in accordance with the filter characteristic determined by the filter control section 105. The filter processing section 106 thereby applies a strong temporal filter to a region wherein flicker noise is liable to be produced, applies a weak temporal filter to a region wherein the after-image is liable to be produced, and also applies a strong spatial filter to a region wherein a pseudo edge is liable to be produced, at a low computation complexity by using the intra prediction information linked with the characteristic of an original image, thereby realizing flicker reduction without causing the after-image. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to an image decoding method and an image decoding apparatus, and more particularly to an image decoding apparatus that reduces flicker, which is a cause of deterioration in subjective image quality, by performing adaptive filter control using intra prediction information. .

  In recent years, with the increase in the number of terrorism and crimes, video surveillance systems capable of connecting tens to hundreds of surveillance cameras are rapidly spreading in public facilities and the like. In addition, since the video from the surveillance camera is used as a criminal evidence image, the image quality of the still image, especially the resolution, is regarded as important, and there is an effect of increasing the number of camera imaging elements. Is progressing.

  Along with the higher resolution of surveillance cameras, surveillance systems that connect dozens of surveillance cameras can efficiently compress high-resolution video in order to record long-term video with limited recording capacity. There is a demand for an image encoding method that can be realized. Furthermore, in order to randomly access the monitoring video, the current situation is that intra coding is used.

  As an image encoding technique that satisfies these requirements, a JPEG (Joint Picture Experts Group) method, a JPEG 2000 method, or the like is applied to a monitoring system, but further improvement in compression performance is required.

  The recently standardized H.264 / AVC (ISO / IEC14496-10) system is an encoding that boasts a compression performance of 1.5 to 2 times that of the conventional MPEG-2 or MPEG-4 video encoding system. It is a standard. High compression performance of H.264 / AVC coding is achieved by improving efficiency of 1) Intra prediction method using intra-screen (intra) correlation and 2) Variable block motion prediction method using inter-image (inter) correlation. It has been realized. For this reason, it has been confirmed that even when video is encoded using intra coding for still image coding, higher compression performance is realized compared to JPEG and JPEG2000 (ISO / IEC JTC1 / SC29). / WG11 MPEG2003 / M10246).

  Therefore, H.264 / AVC intra coding of the surveillance video results in high image quality because each image has high compression performance. However, when it is played back as video, flicker noise is caused by large variations in image quality between images. The problem is that subjective image quality is lowered.

  Conventionally, as a method for reducing flicker noise in H.264 / AVC, for example, in Non-Patent Documents 1 and 2, coding / decoding into a predicted image that is not required in the standard in intra prediction that is a cause of flicker By performing processing (defined as W calculation), the influence of the predicted image is minimized and flicker is reduced.

  FIG. 7 shows an image decoding apparatus described in Non-Patent Document 1.

  In FIG. 7, an image decoding apparatus 700 includes an input unit 701, an entropy decoding unit 702, an inverse quantization orthogonal transformation unit 703, an intra compensation unit 704, an inverse quantization orthogonal transformation unit 706, and an orthogonal transformation quantum. It has a configuration having a conversion unit 705 and an output unit 707.

  In FIG. 7, the input unit 701 outputs the compression-encoded video stream to the entropy decoding unit 702, and the entropy decoding unit 702 performs entropy decoding on the input video stream, thereby orthogonally transforming the intra prediction information. The coefficient is generated, the intra prediction information is output to the intra compensation unit 704, and the orthogonal transform coefficient is output to the inverse quantization orthogonal transform unit 703.

  The inverse quantization orthogonal transform unit 703 performs inverse quantization and orthogonal transform processing on the input orthogonal transform coefficient, generates an intra residual image, and outputs the intra residual image to the intra compensation unit 704.

  The intra compensation unit 704 generates a prediction image using the input intra prediction information with respect to the decoded image for intra prediction stored in the internal memory, and generates an input intra residual image and a prediction image. A decoded image is generated and output to the output unit 707 and the orthogonal transform quantization unit 705. Further, the intra compensation unit updates the image in the internal memory using the decoded image for intra prediction input from the inverse quantization orthogonal transform unit 706.

  The orthogonal transform quantization unit 705 performs orthogonal transform, quantization processing, inverse quantization processing, and orthogonal transform processing on the input decoded image to generate a decoded image for intra prediction, and an intra compensation unit 704. Output to.

  Further, in Patent Document 2, flicker is reduced by using a median filter to suppress luminance fluctuation using a median value of three consecutive frames.

  FIG. 13 shows an image decoding apparatus described in Patent Document 1.

  In FIG. 13, the flicker reduction apparatus 700 includes an input unit 701, a temporal median filter unit 702, an addition unit 703, a spatial median filter unit 704, a coring filter 705, an addition unit 706, and an output unit 707. It has a configuration.

  In FIG. 13, an input unit 701 inputs an image to adders 703 and 706 and a time median filter unit 702. The temporal median filter 702 has two stages of delay circuits, generates a temporal median image from which flicker noise, which is temporally high frequency, is eliminated by replacing pixels using the median value of three consecutive images, and an adding unit 703 Output to.

  The adding unit 703 outputs a flicker image including flicker noise obtained by performing a difference process on the temporal median image from the input image to the spatial median filter unit 704.

  The spatial median filter unit 704 performs a spatial median process on the flicker image, and outputs a flicker spatial median image from which spatial high frequency is eliminated to the coring filter unit 705.

  The coring filter unit 705 eliminates only the high amplitude component of the flicker spatial median image, generates a flicker spatial median image of only the low amplitude component, and outputs the flicker spatial median image to the adding unit 706.

The adder 706 generates an image with reduced flicker by performing a difference process on the flicker space median image having a low amplitude component from the input image, and outputs the generated image to the output unit 707.
Iguchi, "Intra-mode flicker reduction method in 2003 J-040 H.264 encoding", Information Science and Technology Forum 2003 Proceedings, Information Science and Technology Forum, 2003, P277-278 Xiaopeng Fan, “JVT-E070 Flicking Reduction in All Intra Frame Coding”, [online], JVT of ISO / IEC MPEG & ITU-T VCEG, October 2002, [Search February 2, 2005], Internet <URL: http://ftp3.itu.ch/av-arch/jvt-site/2002_10_Geneva/> Special table 2004-503187 gazette

  However, in the non-patent document 1 described above, in order to correctly decode, it is necessary to perform W calculation processing (orthogonal transform quantization unit 705 and inverse quantization orthogonal transform unit 706) that is not originally required in the standard. There is a problem that the amount of calculation increases. Of course, since W calculation is required, it is out of the H.264 / AVC standard.

  Further, in Patent Document 1 described above, since the entire image median filter processing is performed, there is a problem that the amount of calculation is high, and afterimage noise occurs in a scene with a large motion.

  The present invention has been made in view of such points, and by performing post-filtering control adaptively using intra prediction information in a stream, it is possible to reduce flicker noise that has a low calculation amount and no afterimage generation. Objective.

  An image decoding apparatus according to the present invention is an image decoding apparatus for decoding a compression-encoded image stream, and includes entropy decoding means for decoding intra prediction information from the image stream, and the entropy decoding Filter control means for determining a temporal or spatial filter using the intra prediction information decoded by the means, decoding means for decoding the image stream to generate a decoded image, and the time or the time determined by the filter control means Filter processing means for performing a filtering process on the decoded image using a spatial filter.

  According to this configuration, since it is possible to perform filter control using an image stream, it is possible to perform adaptive filter control on a decoded image with a low amount of calculation without the need for a calculation amount for filter control. Is possible.

  Further, the filter control means of the image decoding apparatus according to the present invention determines whether to apply a space or a temporal filter using a prediction block size in the intra prediction information. It determines the characteristics of the time filter.

  According to this configuration, since the filter is controlled according to the block size, it is possible to realize the filter control according to the characteristics of the input image with a low calculation amount.

  The filter control means of the image decoding apparatus according to the present invention determines that a temporal filter is applicable when the prediction block size is larger than a predetermined size in the intra prediction information.

  According to this configuration, for areas where the block size is large and the input image is flat, flicker can be adaptively reduced by applying a temporal filter to flicker noise that reduces subjective image quality. is there.

  The filter control means of the image decoding apparatus according to the present invention determines that the spatial filter is applicable when the prediction block size is smaller than a predetermined size in the intra prediction information.

  According to this configuration, by applying a spatial filter to a pseudo edge noise that is an indirect cause of flicker, for a region where the prediction block size is small and the texture of the input image is fine, pseudo edges are reduced. It can be reduced.

  Further, the filter control means of the image decoding apparatus of the present invention determines whether or not the temporal filter is applicable by using temporal changes between the current frame and the past frame of the intra prediction information, and determines that the temporal filter is applicable. If so, the characteristics of the time or spatial filter are determined.

  According to this configuration, by controlling the filter characteristics based on the temporal change of the intra prediction information that is linked to the temporal change of the input image, it is possible to process the filter process according to the temporal variation of the input image with a low calculation amount. It is.

  The filter control means of the image decoding apparatus according to the present invention determines that the temporal filter is not applied when the prediction block size of the intra prediction information changes with time.

  According to this configuration, the temporal filter is not applied in the region where the temporal change occurs in the input image, and the residual image is generated by the temporal filter in the region where the residual image is likely to occur in the input image. This can be prevented beforehand.

  Further, the filter control means of the image decoding device according to the present invention may reduce the strength of the temporal filter from the current strength or the time when the intra prediction direction of the intra prediction information changes with time. The filter is determined not to be applied.

  According to this configuration, in the region where the texture in the input image changes with time, the strength of the temporal filter is weakened. It is possible to prevent the occurrence of afterimages associated with.

  Further, the filter control means of the image decoding apparatus according to the present invention controls the strength of the temporal or spatial filter using the spatial change of the intra prediction information in the current block and the neighboring blocks.

  In addition, the filter control means of the image decoding apparatus according to the present invention determines the strength of the spatial filter using spatial changes in the current block and the neighboring blocks of the intra prediction information.

  According to this configuration, by controlling the filter characteristics based on the spatial change of the intra prediction information linked to the spatial change of the input image, it is possible to process the filter process according to the spatial variation of the input image with a low calculation amount. It is.

  Further, the filter control means of the image decoding apparatus according to the present invention may reduce the strength of the spatial filter below the current strength when there is a spatial change in the intra prediction direction in the intra prediction information, or The filter is determined not to be applied.

  According to this configuration, the spatial filter is weakened around the edge of the input image where there is a spatial change in the intra prediction information linked to the spatial change of the input image, thereby accompanying the spatial filter around the edge existing in the input image. It is possible to prevent the occurrence of blur.

  The filter control means of the image decoding apparatus according to the present invention determines a spatial filter in a direction perpendicular to the intra prediction direction in the intra prediction information when the spatial filter is determined to be applicable. is there.

  According to this configuration, by controlling the spatial filter for each direction of intra prediction, it is possible to effectively suppress pseudo edges that are likely to occur due to intra prediction.

  Further, the filter control unit of the image decoding device according to the present invention uses a prediction block size in the intra prediction information, compares it with an average prediction block size of neighboring blocks spatially or temporally adjacent, When the predicted block size deviates from the average predicted block size, the application of the spatial or temporal filter is determined after the predicted block size is changed to the average predicted block size of surrounding blocks.

  According to this configuration, it is possible to perform filter control in consideration of temporal or spatial variation of intra prediction information.

  Further, the filter control unit of the image decoding device according to the present invention uses a prediction block size in the intra prediction information, compares it with a spatial prediction block size of a neighboring block spatially or temporally, For a block whose predicted block size deviates from the average predicted block size, the predicted block size of the block is similar to the intra predicted direction of neighboring blocks spatially or temporally adjacent. After changing the prediction block size, the application of the spatial or temporal filter is determined.

  According to this configuration, it is possible to perform filter control in which variation in intra prediction information is taken into consideration and generation of a sense of discomfort due to variation in filter characteristics in peripheral blocks is suppressed.

  The filter control unit of the image decoding apparatus according to the present invention linearly complements the strength of the space or time filter in the boundary region where the application or non-application of the space or time filter is switched.

  According to this configuration, it is possible to suppress deterioration in subjective image quality caused by discontinuity of filter characteristics spatially or temporally at block boundaries.

  An image decoding method according to the present invention is an image decoding method executed by an image decoding device that decodes a compression-encoded image stream, and decodes intra prediction information from the image stream. An entropy decoding step, a filter control step for controlling a temporal or spatial filter using the intra prediction information decoded by the entropy decoding step, and a decoding step for decoding the image stream to generate a decoded image And a filter processing step for performing filter processing on the decoded image using the time or space filter determined by the filter control step.

  According to the present invention, by controlling the characteristics of a temporal / spatial filter applied to a decoded image using intra prediction information obtained by decoding an encoded stream and indicating the image characteristics of an input image, flicker Since it is possible to apply a filter having characteristics suitable for the image characteristics that are the generation factors, it is possible to realize flicker reduction with a low calculation amount and no afterimage generation.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

(Embodiment 1)
FIG. 1 is a block diagram showing a configuration of an image decoding apparatus according to Embodiment 1 of the present invention.

  An image decoding apparatus 100 illustrated in FIG. 1 includes an entropy decoding unit 102, an inverse quantization / orthogonal transformation unit 103, an intra compensation unit 104, a filter control unit 105, and a filter processing unit 106. An input unit 101 and an output It has a configuration connected to the unit 107.

  The input unit 101 is connected to a network or a recording medium, and outputs an encoded stream compressed and encoded by the image encoding device to the entropy decoding unit 102 for each frame.

  The entropy decoding unit 102 entropy-decodes the encoded stream input from the input unit 101 to generate intra prediction information and encoding coefficient information for each block, and the intra prediction information is filtered by the filter control unit 105 and the intra compensation unit 104. And the encoded coefficient information is output to the inverse quantization / orthogonal transform unit 103.

  The inverse quantization / orthogonal transformation unit 103 performs inverse quantization / orthogonal transformation processing on the coding coefficient information input from the entropy decoding unit 102 to generate a decoded residual image that is a residual image of intra prediction. Output to the intra compensation unit 104.

  The intra compensation unit 104 uses the decoded image stored in the internal memory based on the decoded residual image input from the inverse quantization / orthogonal transform unit 103 and the intra prediction information input from the entropy decoding unit 102. The intra compensation process is performed to generate a decoded image, and the decoded image is output to the filter processing unit 106.

  The filter control unit 105 determines the characteristics of the spatial / temporal filter based on the intra prediction information input from the entropy decoding unit 102, and outputs the determined filter characteristics to the filter processing unit 106.

  The filter processing unit 106 performs spatial / temporal filter processing on the decoded image input from the intra compensation unit 104 using the filter characteristics input from the filter control unit 105, and outputs the decoded image after filtering. Output to the output unit 107.

  Next, the operation of the image decoding apparatus 100 having the above configuration will be described using the flowchart shown in FIG.

  2 is stored as a control program in a storage device (for example, a ROM or a flash memory) (not shown) of the image decoding device 100, and is controlled by a CPU (not shown).

<Step S201: Stream Input Process>
First, the input unit 101 is connected to a network or a recording medium, and outputs an encoded stream compressed and encoded by the image encoding device to the entropy decoding unit 102 for each frame.

<Step S202: Entropy Decoding Process>
Next, the entropy decoding unit 102 performs entropy decoding on the encoded stream input from the input unit 101, generates intra prediction information and encoding coefficient information for each block, and converts the intra prediction information between the filter control unit 105 and the intra prediction information. The data is output to the compensation unit 104, and the encoded coefficient information is output to the inverse quantization / orthogonal transform unit 103.

In the present embodiment, H.264 is used as the encoding method of the encoded stream. H.264 / AVC is assumed, and the intra prediction information for each block referred to here is composed of the following two.
(1) Prediction block size A processing unit for performing intra prediction processing. There are two types of 16 × 16 blocks or 4 × 4 blocks.
(2) Intra prediction direction In an intra prediction process, a prediction direction is shown. The direction is different for each block size, and there are 4 directions for 16 × 16 blocks and 9 directions for 4 × 4 blocks.

<Step S203: Stream Decoding Process>
The inverse quantization / orthogonal transformation unit 103 performs inverse quantization / orthogonal transformation processing on the encoded coefficient information input from the entropy decoding unit 102 for each 4 × 4 block, and obtains a decoding residual for the entire frame. An image is generated, and the decoded residual image is output to the intra compensation unit 104.

  The intra compensation unit 104 uses the decoded image stored in the internal memory based on the decoded residual image input from the inverse quantization / orthogonal transform unit 103 and the intra prediction information input from the entropy decoding unit 102. Then, intra compensation processing is performed for each 4 × 4 or 16 × 16 block, a decoded image is generated, the decoded image is output to the filter processing unit 106, and at the same time, the internal memory is updated using the generated decoded image To do.

  3, 4 and 5 are conceptual diagrams showing intra prediction information and intra compensation.

  In FIG. 2, 301 represents a frame, and a 16 × 16 pixel square lattice block is a 16 × 16 block 302 in the frame, and a 4 × 4 pixel square lattice block is within the 16 × 16 block. This is a 4 × 4 block 303.

  In FIG. 3, 401 indicates an intra prediction direction of 4 × 4 blocks, and there are 9 directions in total from 0 to 8. Reference numeral 402 denotes a 4 × 4 block being decoded, 404 denotes a decoding block on the left side, and 403 denotes a decoding block on the upper side.

  In intra-prediction coding, using the decoded pixel values in adjacent blocks A to L, the pixels of a to p are expanded by extending the pixels in one prediction direction obtained by entropy decoding in nine directions. The amount of information is reduced by generating a 4 × 4 predicted image block, performing difference processing with the 4 × 4 block of the original image, and encoding only the remaining blocks.

  In intra-compensation processing in decoding, similarly to intra prediction, a 4 × 4 prediction image block is generated using a prediction direction and adjacent pixels, and a decoded image is acquired by performing addition processing on the 4 × 4 residual block. .

  Note that the second prediction direction in 401 does not indicate a direction, but a predicted image block is generated using an average value of all the adjacent pixels A to L.

  In FIG. 4, 501 indicates the prediction direction of 16 × 16 blocks, and there are a total of 4 directions from 0 to 3. Reference numeral 502 denotes a 16 × 16 block being decoded, reference numeral 504 denotes a decoding block on the left side, and reference numeral 503 denotes a decoding block on the upper side.

  In intra prediction and intra compensation, similarly to the processing of 4 × 4 blocks, adjacent pixels with diagonal lines are expanded in the prediction direction to generate a difference image block, and difference or addition processing is performed.

  In the prediction direction 501, the prediction directions No. 2 and No. 3 do not indicate directions, and the prediction image is obtained using the average value of the whole or the average value of the two diagonal pixels using diagonally adjacent pixels. Is to be generated.

<Step S204: Filter Control Unit>
The filter control unit 105 determines the characteristics of the spatial / temporal filter based on the intra prediction information input from the entropy decoding unit 102, and outputs the determined filter characteristics to the filter processing unit 106.

  FIG. 6 shows the flow of the filter control process.

  First, the filter control unit 105 performs a process of determining whether the block is a 16 × 16 block using the prediction block size of the intra prediction information.

  If the block is 16 × 16 block (corresponding to YES in step S601), the process moves to step S602 related to time filter control. If the block is not 16 × 16 block (in this embodiment, 4 × 4 block) Corresponding to the case (corresponding to NO in step S601), the process moves to step S604 related to the spatial filter control (step S601).

  Next, when it is determined that the block is 16 × 16 block, the temporal filter is controlled using the temporal change of the intra prediction information (steps S602 and S603). Here, the temporal change of the intra prediction information refers to a case where the prediction block size is changed by comparing the intra prediction information in the block at the same location one frame before with the intra prediction information in the current block.

  If there is a change in the predicted block size (corresponding to YES in step S602), the time filter is determined to be OFF, and if there is no change (corresponding to NO in step S602), the process is determined as a time filter strength determination process. Move to (Step S603) (Step S602).

  In this way, when the prediction block size changes with time, it is highly likely that the texture has changed greatly between frames. By turning off the time filter, the time filter is matched to the characteristics of the original image. It is possible to perform noise reduction that suppresses generation of afterimages, which is a problem of the above.

  Next, the strength of the time direction filter is determined using the temporal change of the intra prediction information (step S603). Here, the temporal change in the intra prediction information indicates a case where the prediction direction in the block at the same location one frame before is compared with the prediction direction in the current block, and the prediction direction changes.

  When there is no change in the prediction direction, the strength of the temporal filter is set to be higher than the current strength, and when there is a change in the prediction direction, the strength of the temporal filter is set to be weaker than the current strength.

  The strength of the time filter can be controlled using a filter tap coefficient or the like. For example, Formula 1 shows an example of a time filter.

  In Equation 1, F (X, t) is a pixel value at a coordinate X and a frame number t, F ′ is a pixel value after filtering, X is a coordinate, t is a frame number, and m ( (Corresponding to the tap coefficient) is a filter strength that satisfies K> m. K is a normalization parameter.

  In Equation 1, when the time filter strength is increased, m is set large, and when it is weakened, the time filter strength can be controlled by setting m small. In particular, when the intensity is weakened to the maximum (when m = 0 in Equation 1), it is the same as when the time filter is not applied.

  The time filter strength control method is not limited to the above method, and any method can be used as long as the strength can be controlled.

  Thus, when the prediction direction changes with time, there is a high possibility that a moving object has appeared. By reducing the strength of the time filter, it is possible to reduce noise while suppressing the occurrence of afterimages.

  In addition, when the prediction block size is large and the prediction direction is not temporally converted, the input image is flat and stationary, and the occurrence probability of flicker is high, so by applying a strong temporal filter, It is possible to reduce flicker efficiently. Moreover, the determination process of the temporal change in the prediction direction can be realized with a small amount of calculation that compares the sizes.

  Next, the operation of the filter processing unit 106 when it is determined in step S601 that it is not a 16 × 16 block (that is, equivalent to a 4 × 4 block) will be described (steps S604 and S605). First, spatial change determination processing of intra prediction information is performed (step S604). Here, the spatial filter is turned off. That is, the prediction direction of the surrounding block of the 4 × 4 block is compared with the prediction direction of the current 4 × 4 block. If the prediction direction is different from the surrounding direction (corresponding to YES in step S604), the spatial direction Turn off the filter.

  Note that the amount of change in the prediction direction of the surrounding blocks can be used to gradually decrease the strength of the filter in the spatial direction as the amount of change increases, and finally be set to OFF. For example, the filter strength in the spatial direction can be controlled using the cut-off frequency of the one-dimensional low-pass filter.

  In intra prediction, in encoding prediction direction information, difference information with the prediction direction of the neighboring blocks is encoded, so that a prediction direction similar to the neighboring blocks is easily determined. For this reason, if it is different from the surrounding prediction direction, it is highly possible that the original image contains the original edge. Therefore, it is possible to prevent edge blurring by turning off the spatial filter. is there. Conversely, if there is no spatial change (corresponding to NO in step S604), the process moves to step S605.

  The direction of the spatial filter is determined (step S605). Specifically, a direction perpendicular to the prediction direction is determined as the application direction of the spatial filter using the prediction direction in the intra prediction information. For example, in the 301 prediction direction of FIG. 3, when the prediction direction is the first horizontal prediction, a vertical one-dimensional low-pass filter is used as the spatial filter, and when the prediction direction is 0, the spatial filter Assuming that a horizontal one-dimensional low-pass filter is used.

  In intra prediction, the generation direction of pseudo edges, which are coding noise, is biased depending on the prediction direction, so that the pseudo edges can be efficiently reduced by applying a one-dimensional filter perpendicular to the prediction direction. Is possible. Note that the spatial filter is not limited to a one-dimensional filter having a direction, and any filter can be used as long as it can reduce pseudo edges generated in the prediction direction.

  The above is the processing flow of filter control. In this way, the filter control unit 105 determines the filter characteristics (time / space filter application / non-application information, filter strength information) for all blocks, and outputs the determined filter characteristics to the filter processing unit 106. .

  The filter control unit 105 of the image decoding apparatus 100 calculates an average block size of neighboring blocks spatially or temporally adjacent to each other, and compares it with the average block size. It is also possible to determine the application of the spatial or temporal filter after changing the block size using the block size of blocks having similar intra prediction directions of neighboring blocks spatially or temporally adjacent. Accordingly, it is possible to perform filter control in which variation in intra prediction information is taken into consideration and generation of a sense of discomfort due to variation in filter characteristics in peripheral blocks is suppressed. For example, the similarity in the prediction direction can be determined based on whether or not the mode number value indicating the 401 intra prediction direction in FIG. 4 is close.

  Furthermore, the filter control unit 105 of the image decoding apparatus 100 can also linearly complement the filter strength in the boundary region where the on / off of the spatial or temporal filter is switched. Thereby, it is possible to suppress the deterioration of the subjective image quality caused by the discontinuity of the filter characteristics spatially or temporally. For example, as a linear interpolation method of the filter strength, there is a method of gradually changing the value of m representing the strength in Formula 1 to 0 at the boundary where the filter changes from ON to OFF.

  Subsequently, filter processing is performed (step S205). The filter processing unit 106 uses the filter characteristics input from the filter control unit 105 and the image one frame before stored in the internal memory to perform the time / space on the decoded image input from the intra compensation unit 104. A filter is applied, the decoded image after filtering is output to the output unit 107, and the decoded image before filtering is stored in the internal memory.

  Finally, end determination is performed (step S207). That is, it is determined whether or not the encoded streams of all the images have been processed. If all the encoded streams have been processed, the processing ends. If not, the processing moves to step S201.

  The above is the description of the operation of the image decoding apparatus according to the present embodiment.

  As described above, in the present embodiment, the filter control unit 105 of the image decoding apparatus 100 uses the intra prediction information to control the filter characteristics of the temporal or spatial filter applied to the decoded image, so that the original image Therefore, it is possible to determine a filter characteristic for noise reduction that meets the above characteristics with a low amount of computation.

  In the present embodiment, the filter control unit 105 of the image decoding apparatus 100 controls the filter according to the block size by determining the application of the spatial or temporal filter using the prediction block size of the intra prediction information. Therefore, filter control according to the flatness of the input image can be realized with a small amount of computation.

  In the present embodiment, the filter control unit 105 of the image decoding apparatus 100 applies a temporal filter when the prediction block size of intra prediction information is large (corresponding to 16 × 16 blocks in the present embodiment). By applying a temporal filter to the area where the block size is large, the input image is flat, and flicker that reduces subjective image quality is likely to occur, the flicker can be reduced adaptively. It is.

  In the present embodiment, the filter control unit 105 of the image decoding apparatus 100 applies a spatial filter when the prediction block size is small in the intra prediction information (corresponding to 4 × 4 blocks in the present embodiment). By reducing the pseudo edge by applying a spatial filter to the pseudo edge noise, which is an indirect cause of flicker, for areas where the block size is small and the texture of the input image is fine Is possible.

  In the present embodiment, the filter control unit 105 of the image decoding apparatus 100 controls temporal or spatial filter strength by using temporal changes in intra prediction information in the current frame and past frames. By controlling the filter characteristics based on the temporal change of the intra prediction information that is linked to the temporal change of the input image, it is possible to process the filter process according to the temporal variation of the input image with a low calculation amount. .

  In the present embodiment, filter control section 105 of image decoding apparatus 100 determines that the temporal filter is OFF when the block size in the intra prediction information in the current frame and the past frame changes with time. As a result, it is possible to prevent the occurrence of afterimages in an area where a temporal change occurs in the input image and afterimages are likely to be generated by the time filter.

  In the present embodiment, filter control section 105 of image decoding apparatus 100 weakens the strength of the temporal filter when the intra prediction direction in the intra prediction information in the current frame and the past frame changes temporally. As a result, it is possible to prevent the afterimage from occurring in a region where the texture in the input image changes with time and the afterimage is likely to be generated by the time filter.

  Further, in the present embodiment, the filter control unit 105 of the image decoding apparatus 100 uses the spatial change of the intra prediction information in the current block and the neighboring blocks to control the temporal or spatial filter strength, Since the spatial change of the intra prediction information linked to the spatial change of the input image is used, it is possible to process the filter processing according to the spatial variation of the input image with a low calculation amount.

  In the present embodiment, the filter control unit 105 of the image decoding apparatus 100 weakens the strength of the spatial filter when there is a spatial change in the intra prediction direction in the intra prediction information in the current block and the neighboring blocks. By doing so, the spatial filter is weakened around the edge of the input image where there is a spatial change in intra prediction information, so it is possible to prevent the occurrence of blurring when applying the spatial filter around the original edge existing in the input image It is.

  In the present embodiment, the filter control unit 105 of the image decoding apparatus 100 uses the intra prediction direction in the intra prediction information, and determines the application of the spatial filter in the direction perpendicular to the prediction direction. It is possible to effectively suppress a pseudo edge in a direction that is likely to be generated by prediction.

  Further, in the present embodiment, the filter control unit 105 of the image decoding apparatus 100 uses the block size information in the intra prediction information, and uses the average block size of neighboring blocks spatially or temporally adjacent to each other. Since the application of the spatial or temporal filter is determined after the size is changed, it is possible to perform filter control in which variation in intra prediction information is taken into consideration and generation of a sense of incongruity due to variation in filter characteristics in peripheral blocks is suppressed.

  Further, in the present embodiment, the filter control unit 105 of the image decoding apparatus 100 linearly complements the filter strength in the boundary region where the ON / OFF of the space or the time filter is switched, so that the filter characteristics spatially or temporally. It is possible to suppress the deterioration of the subjective image quality caused by the discontinuity.

  The image decoding apparatus according to the present invention uses intra prediction information obtained from an encoded stream to control the characteristics of a temporal or spatial filter applied to a decoded image, thereby obtaining a filter characteristic that matches the input image characteristic. It is possible to perform calculation with a low amount of computation and reduce flicker without generating afterimages, and is particularly useful in an image decoding apparatus in a surveillance camera system in which the number of cameras is large and high compression is required.

FIG. 1 is a block diagram showing a configuration of an image decoding apparatus according to Embodiment 1. FIG. 7 is a flowchart showing image decoding processing according to the first embodiment. Schematic diagram showing the block size of intra prediction information according to Embodiment 1 The conceptual diagram which shows the prediction direction of 4 * 4 block of the intra prediction information which concerns on Embodiment 1. FIG. The conceptual diagram which shows the prediction direction of the 16 * 16 block of the intra prediction information which concerns on Embodiment 1. FIG. Flow chart showing filter control processing according to Embodiment 1 The block diagram showing the structure of the image decoding apparatus of a prior art Block diagram showing the configuration of a conventional flicker mitigation device

Explanation of symbols

DESCRIPTION OF SYMBOLS 100 Image decoding apparatus 101 Input part 102 Entropy decoding part 103 Inverse quantization and orthogonal transformation part 104 Intra compensation part 105 Filter control part 106 Filter processing part 107 Output part 301 Frame 302 16 * 16 block 303 4 * 4 block 401 4 × 4 block prediction direction 402 4 × 4 block 403,404 Adjacent 4 × 4 block 501 16 × 16 block prediction direction 502 16 × 16 block 503,504 Adjacent 16 × 16 block 700 Image decoding device 701,1301 Input unit 702 Entropy decoding unit 703, 706 Inverse quantization / orthogonal transformation unit 704 Intra compensation unit 705 Orthogonal transformation / quantization unit 707, 1307 Output unit 800 Flicker reduction device 802 Temporal median filter unit 803, 806 Addition unit 804 Sky Median filter unit 805 core filter section

Claims (14)

An image decoding apparatus for decoding a compression-encoded image stream,
Entropy decoding means for decoding intra prediction information from within the image stream;
Filter control means for determining a temporal or spatial filter using intra prediction information decoded by the entropy decoding means;
Decoding means for decoding the image stream to generate a decoded image;
Filter processing means for performing filter processing on a decoded image using the time or spatial filter determined by the filter control means;
An image decoding apparatus. The said filter control means determines the applicability of a space or a temporal filter using prediction block size among the said intra prediction information, and when it determines with applicability, it determines the characteristic of the said spatial or temporal filter. The image decoding apparatus described in 1. The image decoding apparatus according to claim 2, wherein the filter control unit determines that a temporal filter is applicable when a prediction block size of the intra prediction information is larger than a predetermined size. The image decoding device according to claim 2, wherein the filter control unit determines that the spatial filter is applicable when a prediction block size is smaller than a predetermined size in the intra prediction information. The filter control means determines whether or not the temporal filter is applicable using a temporal change between the current frame and the past frame of the intra prediction information, and if it is determined that the temporal filter is applicable, the temporal or spatial filter The image decoding apparatus according to claim 1, wherein the characteristic of the image is determined. The image decoding device according to claim 5, wherein the filter control unit determines that the temporal filter is not applied when a prediction block size of the intra prediction information changes with time. The said filter control means determines that the intensity | strength of the said temporal filter is weakened from the present intensity | strength, or the said temporal filter is unapplied when the intra prediction direction changes temporally among the said intra prediction information. The image decoding apparatus described in 1. The image decoding apparatus according to claim 1, wherein the filter control unit determines the strength of the spatial filter using a spatial change in the current block and the neighboring blocks of the intra prediction information. The filter control means, when there is a spatial change in the intra prediction direction of the intra prediction information, determines that the strength of the spatial filter is weaker than the current strength or that the spatial filter is not applied. The image decoding apparatus according to claim 8. The image decoding device according to claim 4, wherein the filter control unit determines a spatial filter in a direction perpendicular to an intra prediction direction in the intra prediction information when determining that the spatial filter is applicable. The filter control unit compares the average prediction block size of neighboring blocks spatially or temporally using the prediction block size of the intra prediction information, and the current prediction block size is the average prediction block size. 3. The image decoding device according to claim 2, wherein, when deviating, the application of the spatial or temporal filter is determined after the prediction block size is changed to an average prediction block size of neighboring blocks. The filter control unit uses a prediction block size of the intra prediction information to compare with a spatial prediction block size of a neighboring block spatially or temporally, and the current prediction block size determines the average prediction block size. For a deviating block, the prediction direction of the block is changed using the prediction block size of a block similar to the intra prediction direction of neighboring blocks spatially or temporally adjacent, and then the space is changed. Or the image decoding apparatus of Claim 2 which determines application of a time filter. The image decoding apparatus according to claim 2, wherein the filter control unit linearly complements the strength of the space or time filter in a boundary region where application or non-application of the space or time filter is switched. An image decoding method executed by an image decoding apparatus for decoding a compression-encoded image stream,
An entropy decoding step of decoding intra prediction information from within the image stream;
A filter control step for controlling a temporal or spatial filter using the intra prediction information decoded by the entropy decoding step;
A decoding step of decoding the image stream to generate a decoded image;
A filter processing step for performing filter processing on the decoded image using the time or spatial filter determined by the filter control step;
An image decoding method comprising:

Images