JP2015130616A - Image encoding device and program, image decoding device and program, and image encoding system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an image encoding device and an image decoding device capable of improving the quality of a local decoded image or a decoded image by properly executing deblocking filtering.SOLUTION: The image encoding device and the image decoding device each comprises: means which acquires a conspicuousness map and a quantization parameter expressing how much human being momentarily pays attention to each section of the local decoded image or the decoded image; and means which, on the basis of the magnitude of the acquired conspicuousness map and quantization parameter, reviews a block boundary strength calculated once and applies the result to deblocking filtering on the local decoded image or decoded image.

Description

  The present invention relates to an image encoding device and program, an image decoding device and program, and an image encoding system, and is particularly applicable to an image encoding method using a deblocking filter.

  In recent years, with the spread of the Internet, there is an increasing demand for streaming distribution of moving images. In a moving image stream, quality degradation due to subjective evaluation in a limited band of a decoded image due to compression encoding of the moving image is a problem.

  In Non-Patent Document 1, on / off of a deblocking filter can be selected mainly to reduce block noise such as an unnatural color change that occurs at a block boundary.

  A stream encoding process and a stream decoding process described in Non-Patent Document 1 to which a deblocking filter is applied will be briefly described.

  First, the stream encoding process will be described. A prediction method is determined using an input image and a reference image (stored local decoded image after deblocking filter processing) if present, a prediction image is generated from the determined prediction method, and a prediction image is generated from the input image. Subtract to generate residual information. Coefficient information (for example, orthogonal transform coefficient) is generated from the residual information and then encoded (quantized), and a local decoded image (local decoded image) before deblocking filter processing is generated from the encoded information and the predicted image described above. . The deblocking filter process is performed on the local decoded image before the deblocking filter process while referring to the prediction information, and the local decoded image after the process is stored for generation of the next predicted image. Prediction information and encoding information are converted into stream data.

  Next, the stream decoding process will be described. Prediction information is extracted from the input stream data, and a prediction image is generated using the obtained prediction information and a reference image (decoded image after accumulated deblocking filter processing) if present. In addition, the encoding information is extracted from the input stream data and returned to the coefficient information, the residual information is calculated from the obtained coefficient information, and added to the predicted image obtained from the input stream data to be deblocking filtered A decoded image before processing is generated. While referring to the prediction information, the deblocking filter process is performed on the decoded image before the deblocking filter process, and the decoded image after the filter process is stored and used for generation of the next predicted image. , Save, send, print and other output operations.

JP2007-12047A JP2011-39778 Special table 2007-515209 Special table 2010-506315

ITU-U H.I. H.264 / AVC Laurent Itti, Christof Koch, Ernst Niebur, “A Model of Salientity-Based Visual Attention for Rapid Scenario Analysis,” IEEE Transactons Int. 20, no. 11, pp. 1254-1259, November 1998

  However, the technique described in Non-Patent Document 1 has the following problems.

  In the technology described in Non-Patent Document 1, block boundary strength (Bs; Boundary Strength) that defines the strength of a deblocking filter is determined by applying prediction information and quantization parameters of a macroblock including the boundary.

  As described above, since the actual pixel value is not used to determine the block boundary strength, the accuracy of determining whether the block boundary actually has block noise due to the quantization error is low, and the block boundary is strong against the block boundary having weak block noise. In some cases, filtering is performed to remove non-noise components, and conversely, only weak filtering is performed on block boundaries where block noise is strong, and distortion is left conspicuous.

  Therefore, an image encoding device and program, an image decoding device and program, and an image encoding system that can appropriately execute deblocking filter processing and improve the quality of a local decoded image or decoded image are desired.

  A first aspect of the present invention is an image encoding apparatus that includes a deblocking filter that reduces subjective distortion of a block boundary that occurs during encoding of a moving image, and that performs a deblocking filtering process on a local decoded image. Correction judgment information for acquiring correction judgment information corresponding to at least one of a quantization parameter corresponding to a quantization step size according to a predetermined function, and a saliency map indicating a degree of instantaneous attention to each part of an image And (2) a block boundary strength correcting unit that reviews the block boundary strength determined by the deblocking filter based on the size of the acquired correction determination information and applies the deblocking filtering to the local decoded image. It is characterized by having.

  According to a second aspect of the present invention, there is provided an image decoding apparatus including a deblocking filter that reduces subjective distortion of a block boundary that occurs during encoding of a moving image in an opposing image encoding apparatus, and that performs a deblocking filter process on a decoded image (1 ) Acquisition of correction determination information corresponding to at least one of a saliency map representing the degree of instantaneous attention to each part of the decoded image and a quantization step size corresponding to a quantization step size according to a predetermined function Correction determination information acquisition means; and (2) a block boundary to be applied to deblocking filtering for a decoded image by revising the block boundary strength determined by the deblocking filter based on the size of the acquired correction determination information. Strength correction means.

  According to a third aspect of the present invention, there is provided an image encoding system in which an image encoding device that encodes a moving image and an image decoding device that restores a moving image by decoding are opposed to each other. The image coding apparatus according to the present invention is applied, and the image decoding apparatus according to the second aspect of the present invention is applied as the image decoding apparatus.

  An image encoding program according to a fourth aspect of the present invention includes a deblocking filter that reduces subjective distortion at a block boundary that occurs during encoding of a moving image, and is mounted on an image encoding apparatus that performs deblocking filtering on a locally decoded image. The computer corresponds to (1) a saliency map representing the degree of human instantaneous attention to each part of the local decoded image, and a correction corresponding to at least one of quantization parameters corresponding to the quantization step size according to a predetermined function. Correction determination information acquisition means for acquiring determination information; and (2) revising the block boundary strength determined by the deblocking filter based on the size of the acquired correction determination information, Functions as block boundary strength correction means applied to blocking filtering And characterized in that.

  An image decoding program according to a fifth aspect of the present invention includes a deblocking filter that reduces subjective distortion at a block boundary that occurs when a moving image is encoded in an opposing image encoding device, and that performs a deblocking filter process on the decoded image (1) At least one of a saliency map indicating the degree of human attention to each part of the decoded image instantaneously and a quantization parameter corresponding to the quantization step size according to a predetermined function Correction determination information acquisition means for acquiring corresponding correction determination information; (2) based on the size of the acquired correction determination information, the block boundary strength determined by the deblocking filter is reviewed, and the decoded image As a block boundary strength correction means applied to deblocking filtering for Characterized in that to.

  According to the present invention, there are provided an image encoding device and program, an image decoding device and program, and an image encoding system capable of appropriately executing deblocking filter processing and improving the quality of a local decoded image or decoded image. it can.

It is a block diagram which shows the detailed structure of the image coding apparatus of 1st Embodiment. It is a block diagram which shows the detailed structure of the image decoding apparatus of 1st Embodiment. It is a flowchart which shows the review process of the block boundary strength which the deblocking filter main body in the image coding apparatus of 1st Embodiment performs.

(A) First Embodiment Hereinafter, a first embodiment of an image encoding device and program, an image decoding device and program, and an image encoding system according to the present invention will be described in detail with reference to the drawings.

  Here, the image coding apparatus according to the first embodiment is configured by the image coding apparatus according to the first embodiment and the image decoding apparatus according to the first embodiment. In addition, the image encoding method of the first embodiment is H.264. H.264 / AVC.

(A-1) Technology introduced in the first embodiment In the first embodiment, in order to solve the above-described problem, the block boundary strength related to the deblocking filter is used by using a saliency map. To correct it.

  In the field of image processing, a method for calculating bottom-up type attention as an objective index value has been studied. Here, bottom-up attention refers to a process of eye movement that depends on the characteristics of an input image such as color and brightness, which occurs when an observer (person or animal) looks around without any particular purpose. The index value obtained by bottom-up attention is used, for example, in a technique in which a robot instantaneously detects an approach to an object (see Patent Document 1) and a visual content evaluation technique (see Patent Document 2). .

  Non-patent document 2 discloses a saliency map model as a study that models gaze movement by bottom-up attention. The saliency map model is a model that calculates the spatial distribution of saliency in a visual scene for each area in the input image, and predicts eye movement based on the spatial distribution of saliency. The calculation method of the saliency map in the saliency map model is a method of calculating a saliency value obtained by quantifying the degree to which a person instantaneously pays attention to a point or region in an input image, that is, the level of saliency. As applicable.

  There is a region called a receptive field in the retinal ganglion cells in the retina of the eye, and when this receptive field is stimulated by light, the information is transmitted to the brain. The receptive field is composed of a circular center in the center (surround) and its surrounding area (surround). By applying the mechanism in the receptive field, the authors of Non-Patent Document 2 are Center and surround A saliency map was constructed by constructing a model that digitizes the place where the signal becomes strong due to the stimulus (place where attention is drawn).

  Since such a saliency map models human visual characteristics well, it can be expected to detect block noise with high accuracy. Therefore, in the first embodiment, the block boundary strength related to the deblocking filter is corrected (reviewed) using the saliency map to further suppress the block noise.

(A-2) Configuration of First Embodiment FIG. 1 is a block diagram illustrating a detailed configuration of an image encoding device according to the first embodiment. Here, the image encoding device of the first embodiment can be configured by hardware, and can also be realized by software (image encoding program) executed by the CPU and the CPU. Whichever implementation method is employed, it can be functionally represented in FIG.

  In FIG. 1, an image encoding device 10 according to the first embodiment includes an image input unit 11, a prediction residual calculation unit 12, a residual encoding unit 13, a deblocking filter unit 14, a reference image storage unit 15, and a transmission. Part 16. The deblocking filter unit 14 includes a saliency map calculation unit 17 and a deblocking filter body 18.

  The image input unit 11 takes in an input image constituting a moving image and gives it to the prediction residual calculation unit 12.

  The prediction residual calculation unit 12 uses the input image and a reference image (a local decoded image described later) when a reference image exists in the reference image storage unit 15, for example, a prediction method (for example, intra prediction encoding, Inter-screen predictive coding, quantization parameters, etc.) are determined, a predicted image is generated from the input image and / or reference image according to the determined prediction method, and residual information is generated by subtracting the predicted image from the input image Is. The prediction residual calculation unit 12 appropriately divides the input image into macroblocks and blocks at the time of processing. The prediction residual calculation unit 12 gives a prediction image and residual information to the residual encoding unit 13, gives prediction information and residual information to the deblocking filter unit 14, and gives prediction information to the transmission unit 16. Is.

  The residual encoding unit 13 generates coefficient information (for example, an orthogonal transform coefficient) from the residual information, encodes (quantizes) it, and gives the obtained encoded information to the transmission unit 16. The residual encoding unit 13 generates a local decoded image from the predicted image and the encoded information, and supplies the local decoded image to the saliency map calculating unit 17 and the deblocking filter main body 18 in the deblocking filter unit 14.

  The deblocking filter unit 14 performs a deblocking filter process using the saliency map on the local decoded image generated by the residual encoding unit 13, and uses the local decoded image after the filter process as a reference image storage unit 15. It is something to give to.

  The reference image accumulation unit 15 accumulates the filtered local decoded image given from the deblocking filter unit 14 as a reference image that is referred to when a predicted image for the next input image is generated. Note that, depending on the predicted image generation method, the reference image storage unit 15 may store a plurality of filtered local decoded images having different positions on the time axis.

  The transmission unit 16 multiplexes encoded information and prediction information related to an input image, forms a transmission sequence, and performs transmission processing in a broad sense (such as transmission to a network or recording on a recording medium).

  As described above, the deblocking filter unit 14 of the first embodiment is different from that of the conventional image encoding device in that the saliency map is used.

  The saliency map calculation unit 17 of the deblocking filter unit 14 calculates a saliency map using the local decoded image obtained by local decoding obtained by applying the encoding information to the prediction image by the residual encoding unit 13. The calculated saliency map is given to the deblocking filter main body 18. Although detailed description of the calculation method of the saliency map by the saliency map calculation unit 17 is avoided, for example, the calculation method described in any of Patent Documents 2 to 4 and Non-Patent Document 2 can be applied.

  The deblocking filter main body 18 of the deblocking filter unit 14 once obtains the block boundary strength Bs from the prediction information and the local decoded image (local decoded image before filter processing), and then according to the saliency map, Bs is reviewed, and the deblocking filter processing is performed on the local decoded image with reference to the revised block boundary strength Bs, and the local decoded image after the filtering process is obtained and given to the reference image storage unit 15. A method of reviewing the block boundary strength Bs in accordance with the saliency map will be clarified in the description of the operation described later.

  FIG. 2 is a block diagram illustrating a detailed configuration of the image decoding apparatus according to the first embodiment. Here, the image decoding apparatus according to the first embodiment can be configured by hardware, and can also be realized by software (image decoding program) executed by the CPU and the CPU. Even if the realization method is adopted, it can be functionally represented in FIG.

  2, the image decoding apparatus 20 according to the first embodiment includes a stream input unit 21, a predicted image generation unit 22, a residual calculation unit 23, a deblocking filter unit 24, a reference image storage unit 25, and an image output unit 26. Have The deblocking filter unit 24 includes a saliency map calculation unit 27 and a deblocking filter main body 28.

  The stream input unit 21 extracts prediction information and encoding information (quantized coefficient information) from the input stream, provides the prediction information to the prediction image generation unit 22, and provides the coefficient information to the residual calculation unit 23. Is.

  The predicted image generation unit 22 generates a predicted image using the prediction information and the reference image when the reference image is present in the reference image storage unit 25, and gives the predicted image to the residual calculation unit 23 and the deblocking filter unit 24. Is.

  The residual calculation unit 23 restores the residual information from the encoded information (quantized coefficient information), adds the residual information to the predicted image, generates a decoded image, and supplies the decoded image to the deblocking filter unit 24.

  The deblocking filter unit 24 performs a deblocking filter process using the saliency map on the decoded image generated by the residual calculation unit 2, and the decoded image after the filter process is output to the reference image storage unit 25 and the image output. This is given to the unit 26.

  The reference image accumulating unit 25 accumulates the decoded image after the filtering process given from the deblocking filter unit 24 as a reference image that is referred to when generating a predicted image for the next input stream.

  The image output unit 26 performs an output process on the decoded image after the filter process given from the deblocking filter unit 24. The image output unit 26 displays, for example, or records on a recording medium.

  The saliency map calculation unit 27 of the deblocking filter unit 24 calculates a saliency map using the decoded image (pre-filter decoded image) obtained by the residual calculation unit 23, and deciphers the obtained saliency map. This is given to the blocking filter main body 28. The calculation method of the saliency map by the saliency map calculation unit 27 is the same as the calculation method of the saliency map by the saliency map calculation unit 17 in the image encoding device 10.

  The deblocking filter main body 28 of the deblocking filter unit 24 once obtains the block boundary strength Bs from the prediction information and the decoded image (decoded image before filter processing), and then calculates the block boundary strength Bs according to the saliency map. The decoded block image is subjected to deblocking filter processing with reference to the reviewed and reviewed block boundary strength Bs, and the decoded image after the filter processing is obtained and supplied to the reference image storage unit 25 and the image output unit 26. The method of reviewing the block boundary strength Bs according to the saliency map is the same as the review method by the deblocking filter main body 18 in the image encoding device 10.

(A-3) Operation of the First Embodiment Next, the operation of the image encoding device 10 according to the first embodiment will be described.

  When one image constituting a moving image is input to the image encoding device 10, the input image is given from the image input unit 11 to the prediction residual calculation unit 12, and the prediction residual calculation unit 12 inputs the input image. And the reference image are appropriately used to determine the prediction method. Further, the prediction residual calculation unit 12 generates a prediction image from the input image and / or the reference image according to the determined prediction method, and then subtracts the prediction image from the input image to generate residual information. After the coefficient information is generated by transform processing on the residual information in the residual encoding unit 13, the coefficient information is quantized (encoded) in a quantization step corresponding to the quantization parameter, and the obtained encoded information is obtained. Is given to the transmitter 16. Further, in the residual encoding unit 13, a local decoded image is generated (restored) from the predicted image and the encoded information, and is provided to the saliency map calculating unit 17 and the deblocking filter main body 18 in the deblocking filter unit 14. In the transmission unit 16, encoded information and prediction information related to the input image are combined and a transmission sequence is formed and transmitted.

  In the deblocking filter unit 14, a saliency map is calculated by using the local decoded image obtained by local decoding by the saliency map calculation unit 17, and is provided to the deblocking filter main body 18. Then, after the block boundary strength Bs is once determined from the prediction information and the local decoded image by the deblocking filter body 18, the block boundary strength Bs is reviewed according to the saliency map, and the reviewed block boundary strength Bs. And the deblocking filter process is performed on the local decoded image. The local decoded image after the filter processing is given to the reference image storage unit 15 and stored therein.

  The method for reviewing the block boundary strength Bs in the deblocking filter body 18 is as follows. FIG. 3 is a flowchart showing processing by the review method.

  The deblocking filter main body 18 calculates a total value (hereinafter referred to as a total saliency map value) Tmap of all the pixels (pixels) included in the vertical and horizontal 4 × 4 pixel block to be processed (step Smap). S100).

  Next, the deblocking filter body 18 compares the total saliency map value Tsmap with a preset threshold s, and the quantization parameter QP of the macroblock including the 4 × 4 block to be processed is set in advance. Compared with the threshold value t, the 4 × 4 block to be processed is classified (step S101).

  When the total saliency map value Tsmap is large, it means that the 4 × 4 block to be processed belongs to an area to which attention is directed without being conscious of human beings. It represents a big difference. The quantization parameter QP having a predetermined functional relationship with the quantization step size reflects the dynamic range of the residual information in the macroblock, and a large quantization parameter QP means that the dynamic range of the residual information is large and the previous image is large. This indicates that the macro block has a large change.

  In the first embodiment, when the total saliency map value Tsmap is greater than the threshold s and the quantization parameter QP is greater than the threshold t, it is determined that the 4 × 4 block to be processed is a high noise block. To do. When the total saliency map value Tsmap is smaller than the threshold s and the quantization parameter QP is smaller than the threshold t, it is determined that the 4 × 4 block to be processed is a low noise block. Furthermore, when the set of the total saliency map value Tsmap and the quantization parameter QP does not correspond to the two cases described above, it is determined that the 4 × 4 block to be processed is a normal block.

  When the deblocking filter main body 18 determines that the 4 × 4 block to be processed is a normal block, the deblocking filter main body 18 maintains the value of the block boundary strength Bs without correction (step S102).

  When the deblocking filter main body 18 determines that the 4 × 4 block to be processed is a high noise block, the deblocking filter main body 18 increments the value of the block boundary strength Bs by 1 in order to strengthen the deblocking filtering (step S103), and then corrects it. It is determined whether or not the subsequent block boundary strength Bs has exceeded the upper limit of 4 (step S104), and if it has exceeded, the corrected block boundary strength Bs is set to the upper limit of 4 (step S105). .

  If the deblocking filter body 18 determines that the 4 × 4 block to be processed is a low noise block, the deblocking filter main body 18 decrements the value of the block boundary strength Bs by 1 in order to weaken the deblocking filtering (step S106), and then corrects it. It is determined whether or not the subsequent block boundary strength Bs is below the lower limit value of 0 (step S107), and if it is lower, the corrected block boundary strength Bs is set to the lower limit value of 0 (step S108). .

  The deblocking filter process is executed by applying the block boundary strength Bs reviewed (corrected) as described above.

  Next, the operation of the image decoding device 20 according to the first embodiment will be described.

  When the sequence transmitted by the image encoding device 10 according to the first embodiment arrives at the image decoding device 20 according to the first embodiment, the sequence is input to the stream input unit 21. In the stream input unit 21, prediction information and encoding information (quantized coefficient information) are extracted from the input stream, the prediction information is given to the prediction image generation unit 22, and the coefficient information is the residual calculation unit 23. Given to. As a result, the prediction image generation unit 22 generates a prediction image using the prediction information and the reference image when the reference image is present in the reference image storage unit 25, and the residual calculation unit 23 and the deblocking filter The residual calculation unit 23 restores the residual information from the encoded information, adds the residual information to the predicted image, generates a decoded image, and supplies the decoded image to the deblocking filter unit 24.

  In the saliency map calculation unit 27 in the deblocking filter unit 24, the saliency map is calculated by applying the decoded image in the same manner as the saliency map calculation unit 17 in the image encoding device 10, and the deblocking filter main body 28. Given to. In the deblocking filter main body 28, after the block boundary strength Bs is once determined from the prediction information and the decoded image, the block boundary strength Bs is reviewed according to the saliency map, and the reviewed block boundary strength Bs is obtained. The decoded image is referred to and subjected to deblocking filter processing. The review method here is the same as the review method by the deblocking filter main body 18 in the image encoding device 10.

  The decoded image after the deblocking filter processing is given to the reference image storage unit 25 and the image output unit 26. In the reference image storage unit 25, the decoded image after the deblocking filter processing is stored as a reference image that is referred to when a predicted image for the next stream is generated. Also, the decoded image after the deblocking filter process is output from the image output unit 26.

(A-4) Effect of First Embodiment According to the first embodiment, a block boundary that has a high possibility of having noise or noise is detected using a saliency map and a quantization parameter. Since it is possible to determine with high accuracy that the block boundary is unlikely to have, the block boundary strength applied by the deblocking filter is corrected based on the determination result, so the deblocking filter processing is executed properly. Thus, the quality of the local decoded image and the decoded image can be improved by improving the quality of the reference image.

  In the image coding apparatus, since high frequency components added due to unnatural boundary pixel value changes are removed from the reference image, coding efficiency can be improved.

  In the image encoding device and the image decoding device, the same saliency map calculation method is used for the local decoded image and the code image before the deblocking filter processing, and thus the same applies to the image encoding device and the image decoding device. The saliency map can be calculated, and the same contents can be obtained as the local decoded image and the decoded image.

(B) Other Embodiments The present invention is not limited to the above-described first embodiment, and various modified embodiments exemplified below can be given.

  In the first embodiment, the block boundary strength Bs is incremented by 1 or decremented by 1 when the block boundary strength Bs is reviewed. However, the change width is not limited to 1 and may be 2 or more. In this case, at least one of a threshold value for the composite saliency map value and a threshold value for the quantization parameter is provided, and the case classification may be varied, for example, when the change width is set to 1 or 2. good. Furthermore, the change width may be changed in the change direction, for example, +2 when the block boundary strength Bs is increased and −1 when the block boundary strength Bs is decreased.

  Further, the change of the block boundary strength Bs may be changed to a value determined in advance as the changed block boundary strength Bs, instead of a relative method of increasing or decreasing the change width. For example, it is necessary to prepare a table in which the block boundary strength Bs is described in association with the combination of the range to which the combined saliency map value belongs and the range to which the quantization parameter belongs, and to change the block boundary strength Bs once determined. In this case, a new block boundary strength Bs may be obtained by referring to the table.

  In the first embodiment, the case where the changed block boundary strength Bs is outside the range of 0 to 4 has been shown to be returned to the range. However, the deblocking filter process may be performed without returning. For example, an FIR filter or the like for the case where the block boundary strength Bs is 5 may be determined, and if the block boundary strength Bs is 5, the FIR filter for that may be applied to perform filtering. Further, when the block boundary strength Bs is −1, it may be handled that there is no filtering without returning to 0.

  In the first embodiment, the threshold value s for the combined saliency map value is a threshold value that functions both when the block boundary strength Bs is increased and when it is decreased. The threshold value sd for determining whether or not to decrease may be different values. A similar modification can be given for the threshold value for the quantization parameter.

  In the first embodiment, the threshold value s to be compared with the total saliency map is a predetermined value, but may be set dynamically. For example, different values may be applied as the threshold s depending on whether the macro block to which the 4 × 4 block to be processed belongs is an intra macro block or an inter macro block.

  Similarly, the threshold t to be compared with the quantization parameter QP may be different depending on whether the macro block to which the 4 × 4 block to be processed belongs is an intra macro block or an inter macro block.

  In the first embodiment, the case where the increase or decrease in the review of the block boundary strength Bs is avoided has been described. However, only the increase may be performed in the review.

  In the first embodiment, the case where the block size for reviewing the block boundary strength Bs is 4 × 4 pixels in the vertical and horizontal directions. However, other sizes may be used for the block size for reviewing the block boundary strength Bs. You may make it. For example, an 8 × 8 block or an 8 × 16 block may be applied, and the shape is not limited to a rectangle.

  In the first embodiment, the case where the calculation method of the saliency map value fixed in advance by the image encoding device and the image decoding device is shown, but a plurality of saliency map value calculation methods can be applied, You may make it selectable. For example, the image encoding device may select a calculation method according to an operation of a user or the like, and include information on the selected calculation method in a stream and notify the decoding device. The calculation method applied in the negotiation performed in order to match the parameters and the like may be adjusted between the communication device including the image encoding device and the communication device including the image decoding device.

  In the first embodiment, the block boundary strength Bs is reviewed by applying both the combined saliency map value and the quantization parameter. However, the block boundary strength Bs may be reviewed by applying at least one of them. .

  In the first embodiment, the case where both the image encoding device and the image decoding device have the function of reviewing the block boundary strength Bs is shown, but only at least one of the image encoding device and the image decoding device is provided. May be.

  The first embodiment is described in H.264. An image encoding device and an image decoding device compliant with H.264 / AVC have been shown, but MPEG-H HEVC / H. The present invention can also be applied to an apparatus that complies with an image coding system having a similar filter such as H.265.

10: Image encoding device,
DESCRIPTION OF SYMBOLS 11 ... Image input part, 12 ... Prediction residual calculation part, 13 ... Residual encoding part, 14 ... Deblocking filter part, 15 ... Reference image storage part, 16 ... Transmission part, 17 ... Saliency map calculation part, 18 ... deblocking filter body,
20 Image decoding device,
DESCRIPTION OF SYMBOLS 21 ... Stream input part, 22 ... Predictive image generation part, 23 ... Residual calculation part, 24 ... Deblocking filter part, 25 ... Reference image storage part, 26 ... Image output part, 27 ... Saliency map calculation part, 28 ... Deblocking filter body.

Claims (10)

In an image encoding device that includes a deblocking filter that reduces subjective distortion of a block boundary that occurs during encoding of a moving image, and that performs deblocking filtering on a local decoded image,
Acquire saliency map indicating the degree of human instantaneous attention to each part of the local decoded image, and correction determination information corresponding to at least one of quantization parameters corresponding to a quantization step size according to a predetermined function Correction determination information acquisition means;
Block boundary strength correction means for revising the block boundary strength determined by the deblocking filter based on the size of the acquired correction determination information and applying the block boundary strength to the local decoding image. An image encoding device.   The block boundary strength correction unit corrects the block boundary strength upward when the size of the acquired correction determination information is large enough to determine that the block boundary to be processed is likely to have noise. The image coding apparatus according to claim 1, wherein the image coding apparatus is applied to deblocking filtering for a local decoded image.   The block boundary strength correction unit corrects the block boundary strength downward when the size of the acquired correction determination information is large enough to determine that the block boundary to be processed is less likely to have noise. The image coding apparatus according to claim 1, wherein the image coding apparatus is applied to deblocking filtering for a local decoded image. In an image decoding apparatus that includes a deblocking filter that reduces subjective distortion of a block boundary that occurs during encoding of a moving image in an opposing image encoding apparatus, and that performs a deblocking filter process on a decoded image,
Correction for acquiring correction judgment information corresponding to at least one of a saliency map representing the degree of instantaneous attention to each part of a decoded image and a quantization step size corresponding to a quantization step size according to a predetermined function Information acquisition means for determination;
Block boundary strength correction means for reviewing the block boundary strength determined by the deblocking filter based on the size of the acquired correction determination information and applying the block boundary strength to the deblocking filtering on the decoded image. Image decoding device.   The block boundary strength correction unit corrects the block boundary strength upward when the size of the acquired correction determination information is large enough to determine that the block boundary to be processed is likely to have noise. The image decoding apparatus according to claim 4, wherein the image decoding apparatus is applied to deblocking filtering on a decoded image.   The block boundary strength correction unit corrects the block boundary strength downward when the size of the acquired correction determination information is large enough to determine that the block boundary to be processed is less likely to have noise. 6. The image decoding apparatus according to claim 4, wherein the image decoding apparatus is applied to deblocking filtering for a decoded image. In an image encoding system in which an image encoding device that encodes a moving image and an image decoding device that restores the moving image by decoding are opposed to each other,
An image coding system according to claim 1, wherein the image coding device according to claim 1 is applied as the image coding device, and the image decoding device according to claim 4 is applied as the image decoding device.   The correction determination information acquisition unit of the image encoding device calculates a saliency map, and the correction determination information acquisition unit of the image decoding device calculates a saliency map. The image encoding system according to claim 7, wherein the calculation method is the same for the image encoding device and the image decoding device. A computer mounted on an image encoding device that includes a deblocking filter that reduces subjective distortion at a block boundary that occurs during encoding of a moving image, and that performs deblocking filtering on a local decoded image,
Acquire saliency map indicating the degree of human instantaneous attention to each part of the local decoded image, and correction determination information corresponding to at least one of quantization parameters corresponding to a quantization step size according to a predetermined function Correction determination information acquisition means;
Based on the size of the obtained correction determination information, the block boundary strength determined by the deblocking filter is reviewed to function as a block boundary strength correction unit that is applied to deblocking filtering for a local decoded image. A characteristic image encoding program. A computer mounted in an image decoding device that includes a deblocking filter that reduces subjective distortion of a block boundary that occurs when encoding a moving image in an opposing image encoding device, and that performs a deblocking filter process on a decoded image,
Correction for acquiring correction judgment information corresponding to at least one of a saliency map representing the degree of instantaneous attention to each part of a decoded image and a quantization step size corresponding to a quantization step size according to a predetermined function Information acquisition means for determination;
The block boundary strength determined by the deblocking filter is reviewed based on the size of the acquired correction determination information, and the block boundary strength correction means is applied to the deblocking filtering for the decoded image. An image decoding program.

Images