WO2015196860A1 - 一种图像处理方法、装置及系统 - Google Patents

一种图像处理方法、装置及系统 Download PDF

Info

Publication number
WO2015196860A1
WO2015196860A1 PCT/CN2015/077721 CN2015077721W WO2015196860A1 WO 2015196860 A1 WO2015196860 A1 WO 2015196860A1 CN 2015077721 W CN2015077721 W CN 2015077721W WO 2015196860 A1 WO2015196860 A1 WO 2015196860A1
Authority
WO
WIPO (PCT)
Prior art keywords
predicted
pixel
component
image block
image processing
Prior art date
Application number
PCT/CN2015/077721
Other languages
English (en)
French (fr)
Inventor
朱策
马姝颖
林永兵
郑建铧
宋剑军
Original Assignee
华为技术有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 华为技术有限公司 filed Critical 华为技术有限公司
Publication of WO2015196860A1 publication Critical patent/WO2015196860A1/zh

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/50—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding
    • H04N19/597—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding specially adapted for multi-view video sequence encoding

Definitions

  • the present invention relates to the field of multimedia, and in particular, to an image processing method, apparatus, and system.
  • SDC Segment-wise Direct Component Coding
  • 3D-HEVC High Efficiency Video Coding
  • the DLT Depth Lookup Table
  • the video decoding device for example, a decoder
  • the video decoding device indexes the index through the DLT.
  • the residual value is converted into a depth residual value, and finally the depth residual value is added to the pixel value of each pixel in the corresponding region (for example, a prediction image block) to obtain a reconstructed image block.
  • DC Direct Component
  • SDC SDC
  • the original DC Olet DC
  • the predicted DC Reference DC
  • Depth images have very few different pixel values.
  • the number of coded signal bits can be effectively reduced.
  • the average of four pixel values of the four corner pixels of the upper left corner, the upper right corner, the lower left corner, and the lower right corner of the predicted image block is used as the predicted DC.
  • the predictive DC plays a role in assisting in calculating the depth residual value
  • the depth residual value represents the overall deviation of the predicted image block from the original image block.
  • the accuracy of the depth residual difference prediction is directly related to the generation.
  • the coding performance when reconstructing image blocks is good or bad.
  • the DLT transform will generate transform errors, affecting the accuracy of depth residual prediction, and thus affect the coding performance of the reconstructed image block. Therefore, there are some methods to reduce the transform error in order to improve the coding performance when generating reconstructed image blocks.
  • the coding gain is not obtained, and the coding gain reflects the improvement of the coding performance when reconstructing the image block.
  • the existing improved techniques for the conventional method of performing SDC on the depth image cannot reduce the conversion error generated by the DLT, and cannot improve the coding performance when reconstructing the image block of the depth image.
  • Embodiments of the present invention provide an image processing method, apparatus, and system, which can reduce conversion errors generated by DLT when performing SDC image processing on a depth map, thereby improving coding performance of image processing results.
  • an image processing method including:
  • the depth query table includes at least one index value and a pixel value corresponding to each index value, the index value being used to represent a pixel value corresponding to the index value itself;
  • the preferred image processing mode, the index residual value corresponding to the preferred image processing mode, and the depth lookup table are encoded to obtain encoded data.
  • the predicted DC component is obtained according to the pixel value of one pixel in the predicted image block, specifically:
  • the predicted image block central region is a square image region having a side length of 2 pixels at a center of the predicted image block.
  • obtaining the predicted DC component according to the pixel value of one pixel in the predicted image block includes:
  • the image processing mode is an angle mode, acquiring a predicted DC component according to a pixel value of one pixel in the predicted image block;
  • the predicted DC component is obtained according to the pixel value of one pixel in the predicted image block, include:
  • the predicted DC component is obtained according to the pixel value of one pixel in the predicted image block.
  • the predicted DC component is obtained according to the pixel value of one pixel in the predicted image block, include:
  • the predicted DC component is obtained according to the pixel value of one pixel in the predicted image block.
  • acquiring the original DC component according to the original image block includes:
  • the original DC component is obtained from the pixel values of one pixel in the original image block.
  • the acquiring the original DC component according to the pixel value of one pixel in the original image block includes:
  • An original DC component is obtained according to a pixel value of one pixel of the central portion of the original image block, the original image block central region being a square image region having a side length of 2 pixels at the center of the original image block.
  • each image processing mode is acquired according to the at least one image processing mode
  • the predicted image block specifically includes:
  • the predicted image blocks corresponding to each of the image processing modes are acquired according to the at least one image processing mode.
  • the at least one image processing is performed according to the predicted image block and a preset rule
  • the preferred image processing mode is selected in the mode, and specifically includes:
  • An image processing mode that minimizes the rate distortion penalty in the at least one image processing mode is a preferred image processing mode according to the rate distortion penalty.
  • an image processing method including:
  • the encoded data includes an image processing mode, an index residual value, and a depth query table, where the depth query table includes at least one index value and a pixel value corresponding to each index value,
  • the index value is used to represent a pixel value corresponding to the index value itself;
  • the predicted DC component is obtained according to the pixel value of one pixel in the predicted image block, specifically:
  • the predicted image block central region is a square image region having a side length of 2 pixels at a center of the predicted image block.
  • obtaining the predicted DC component according to the pixel value of one pixel in the predicted image block includes:
  • the image processing mode is an angle mode, acquiring a predicted DC component according to a pixel value of one pixel in the predicted image block;
  • obtaining the predicted DC component according to the pixel value of one pixel in the predicted image block includes:
  • the predicted DC component is obtained according to the pixel value of one pixel in the predicted image block.
  • the predicted DC component is obtained according to the pixel value of one pixel in the predicted image block, include:
  • the predicted DC component is obtained according to the pixel value of one pixel in the predicted image block.
  • acquiring the predicted image block according to the image processing mode includes:
  • a video encoding apparatus including:
  • an acquiring unit configured to acquire an original image block of the depth image to be encoded, at least one image processing mode, and a depth query table, where the depth query table includes at least one index value and a pixel value corresponding to each index value, The index value is used to represent the pixel value corresponding to the index value itself;
  • the acquiring unit is further configured to acquire, according to the at least one image processing mode acquired by the acquiring unit, a predicted image block corresponding to each image processing mode;
  • a prediction index value unit configured to obtain a prediction index value according to the depth query table acquired by the acquiring unit and the predicted DC component acquired by the DC component unit.
  • An original DC component unit configured to acquire an original DC component according to the original image block acquired by the acquiring unit
  • an original index value unit configured to obtain an original index value according to the depth query table acquired by the acquiring unit and the original DC component obtained by the original DC component unit;
  • An index residual unit configured to obtain the index residual value by subtracting the original index value obtained by the original index value unit from the predicted index value obtained by the prediction index value unit;
  • a selecting unit configured to select a preferred image processing mode from the at least one image processing mode according to the predicted image block and a preset rule acquired by the acquiring unit;
  • the depth lookup table is encoded to obtain encoded data.
  • the predicted DC component is specifically configured to obtain a predicted DC component according to a pixel value of a pixel in a lower right corner of the predicted image block acquired by the acquiring unit;
  • the predicted DC component unit is specifically configured to acquire a predicted DC component according to a pixel value of one pixel of the central region of the predicted image block acquired by the acquiring unit, where the predicted image block center region is at a center of the predicted image block A square image area with a side length of 2 pixels.
  • the predicted DC component unit is further configured to: when the image processing mode acquired by the acquiring unit is an angle mode, Obtaining a predicted DC component according to a pixel value of one pixel in the predicted image block acquired by the acquiring unit;
  • the predicted DC component unit is further used for acquiring in the acquiring unit
  • the predicted DC component is obtained according to the pixel value of one pixel in the predicted image block acquired by the acquiring unit.
  • the predicted DC component unit is further used for acquiring in the acquiring unit
  • the predicted DC component is obtained according to the pixel value of one pixel in the predicted image block acquired by the acquiring unit.
  • the original DC component unit is specifically used to obtain the Taking the pixel value of one pixel in the original image block obtained by the unit to obtain the original DC component.
  • the original DC component unit is specifically configured to be used in the lower right corner pixel of the original image block acquired according to the acquiring unit.
  • the pixel value obtains the original DC component
  • the original DC component unit is configured to acquire an original DC component according to a pixel value of a pixel of the central region of the original image block acquired by the acquiring unit, where the original image block central region is at a center of the original image block A square image area with a side length of 2 pixels.
  • the acquiring unit is specifically configured to: when the intra-frame reference pixel filter is turned off, according to And the at least one image processing mode acquired by the acquiring unit acquires a predicted image block corresponding to each of the image processing modes.
  • the preferred unit includes:
  • a rate-distortion sub-unit configured to calculate, according to the predicted image block acquired by the acquiring unit, a rate distortion cost of an image processing mode corresponding to the predicted image block in the at least one image processing mode;
  • a preferred subunit configured to use, as the preferred image processing, an image processing mode in which the rate distortion cost of the at least one image processing mode acquired by the acquiring unit is minimized according to the rate distortion cost calculated by the rate distortion subunit mode.
  • a fourth aspect provides a video decoding apparatus, including:
  • An acquiring unit configured to acquire encoded data of a depth image to be reconstructed, where the encoded data includes an image processing mode, an index residual value, and a depth query table, where the depth query table includes at least one index value and each index value corresponds to a pixel value, the index value is used to represent a pixel value corresponding to the index value itself;
  • the obtaining unit is further configured to acquire a predicted image block according to the image processing mode acquired by the acquiring unit;
  • a prediction index value unit configured to obtain a prediction index value according to the depth query table acquired by the acquiring unit and the predicted DC component acquired by the DC component unit.
  • a re-index value unit configured to add the index residual value obtained by the acquiring unit to the prediction index value obtained by the prediction index value unit to obtain a reconstruction index value
  • Reconstructing a DC component unit configured to obtain a reconstructed pixel value according to the depth query table obtained by the acquiring unit and the reconstructed index value obtained by the reconstructed index value unit, Reconstructing pixel values as reconstructed DC components;
  • a depth residual difference unit configured to subtract the reconstructed direct current component obtained by the reconstructed direct current component unit from the predicted direct current component to obtain a depth residual value
  • a reconstruction unit configured to obtain, by using a pixel value of each pixel of the predicted image block acquired by the acquiring unit, the depth residual value obtained by adding the depth residual unit to obtain a reconstructed image block, where the reconstructing The image block is used to reconstruct the depth image to be reconstructed.
  • the predicted DC component is specifically configured to obtain a predicted DC component according to a pixel value of a pixel in a lower right corner of the predicted image block acquired by the acquiring unit;
  • the predicted DC component unit is specifically configured to acquire a predicted DC component according to a pixel value of one pixel of the central region of the predicted image block acquired by the acquiring unit, where the predicted image block center region is at a center of the predicted image block A square image area with a side length of 2 pixels.
  • the predicted DC component unit is further configured to: when the image processing mode acquired by the acquiring unit is an angle mode, Obtaining a predicted DC component according to a pixel value of one pixel in the predicted image block acquired by the acquiring unit.
  • the predicting direct current component unit, and the image processing mode used for acquiring by the acquiring unit is In the planar mode, the predicted DC component is obtained according to the pixel value of one pixel in the predicted image block acquired by the acquiring unit.
  • the predicted DC component unit is further used for acquiring in the acquiring unit
  • the predicted DC component is obtained according to the pixel value of one pixel in the predicted image block acquired by the acquiring unit.
  • the acquiring unit is specifically configured to: when the intra-frame reference pixel filter is turned off, according to The image processing mode acquired by the acquiring unit acquires a predicted image block.
  • the image processing method, device and system provided by the embodiments of the present invention can obtain an original image block of a depth image to be encoded, at least one image processing mode and a depth query table by using a video encoding device, and acquire a predicted image block according to the image processing mode. And obtaining the original DC component according to the obtained information, combining the specific image processing mode to obtain the predicted DC component according to the pixel value of one pixel in the predicted image block, and performing the DLT calculation index according to the original DC component and the predicted DC component.
  • the residual value is used to reconstruct the encoded depth image from the index residual value.
  • the encoded data of the depth image to be reconstructed is obtained by the video decoding device, where the encoded data includes an image processing mode, an index residual value, and a depth query table, and the The specific image processing mode acquires the predicted DC component according to the pixel value of one pixel in the predicted image block, and then performs DLT transformation, and combines the index residual value and the DLT inverse transform to obtain the reconstructed image block, so as to finally obtain the reconstructed image.
  • the probability that a single pixel value is a valid pixel value is much greater than the probability that multiple pixel values are effective pixels, so the SDC encoding process is reduced.
  • the transform error generated by the DLT in the middle improves the coding performance of the image processing result.
  • FIG. 1 is a schematic flowchart diagram of an image processing method according to an embodiment of the present invention
  • FIG. 2 is a schematic diagram of an image processing mode according to an embodiment of the present invention.
  • FIG. 3 is a schematic diagram of application of a deep lookup table according to an embodiment of the present invention.
  • FIG. 4 is a schematic diagram of application of another depth lookup table according to an embodiment of the present invention.
  • FIG. 5 is a schematic flowchart diagram of an image processing method according to still another embodiment of the present invention.
  • FIG. 6 is a schematic flowchart diagram of an image processing method according to another embodiment of the present invention.
  • FIG. 7 is a schematic diagram of acquiring a predicted DC component according to another embodiment of the present invention.
  • FIG. 8 is a schematic flowchart diagram of an image processing method according to still another embodiment of the present invention.
  • FIG. 9 is a schematic structural diagram of a video encoding apparatus according to an embodiment of the present invention.
  • FIG. 10 is a schematic structural diagram of still another video encoding apparatus according to an embodiment of the present invention.
  • FIG. 11 is a schematic structural diagram of a video decoding apparatus according to an embodiment of the present invention.
  • FIG. 12 is a schematic structural diagram of a video encoding apparatus according to still another embodiment of the present invention.
  • FIG. 13 is a schematic structural diagram of a video decoding apparatus according to still another embodiment of the present invention.
  • the embodiment of the invention provides an image processing system, which is applied to the multimedia field, and can be specifically used for encoding processing and decoding processing of a depth image. It can be applied to SDC processing of depth maps in 3D-HEVC.
  • the image processing system includes a video encoding device and a video decoding device that are connectable to each other for data communication.
  • the video encoding apparatus needs to first select the image block to be encoded (such as the original image block) and the reference image block (such as the predicted image block) of the depth image, and then respectively obtain the original DC component of the original image block and Predicting the predicted DC component of the image block, and transforming the predicted DC component with the established DLT (Depth Lookup Table) to obtain a prediction index value, and transforming the original DC component to obtain the original index value, and calculating An index residual value between the original index value and the predicted index value, and encoding the index residual value into the encoded data.
  • the image block to be encoded such as the original image block
  • the reference image block such as the predicted image block
  • the video decoding device decodes the encoded data to obtain the index residual value, and obtains a predicted image block using the same predicted image block selection mode as the video encoding end device, and obtains a predicted DC component, and obtains a prediction in combination with DLT.
  • the index value is added to the index residual value and the predicted index value to obtain a reconstructed index value, and then inversely transformed with the DLT to obtain a reconstructed DC component, and the reconstructed DC component is subtracted from the predicted DC component to obtain a depth residual value.
  • the depth residual value is added to the pixel value of each pixel of the predicted image block to obtain a reconstructed image block.
  • SDC is a residual data encoding method applied in 3D-HEVC.
  • the SDC can be applied in all intra prediction modes of the depth image, and the intra prediction mode is an image processing mode in which the image to be processed is predictively encoded using each frame of the image to be processed, including the HEVC intra prediction mode.
  • the HEVC intra prediction mode includes a DC component mode, a planar mode, and 33 angular modes.
  • each PU (Prediction Unit) in the depth image is taken as a region, and the size of the region is always 2N ⁇ 2N, and the unit is a pixel, and N is an integer greater than 1.
  • Each region (eg, a predictive image block) is encoded by a video encoding device (eg, an encoder) by encoding a DLT (Depth Lookup Table) into an encoded data and sent to a video decoding device (eg, a decoder).
  • the video decoding device further converts the index residual value into a depth residual value through the DLT, and finally adds the depth residual value to the pixel value of each pixel in the corresponding region (for example, the predicted image block) to obtain a reconstructed image block.
  • DC Direct Component, DC component
  • SDC SDC
  • the original DC Olet DC
  • Prediction DC predicted DC
  • Depth images have very few different pixel values.
  • the transformations performed in conjunction with DLT are irreversible.
  • the DC component obtained by the non-effective pixel is indexed by DLT, there is no direct corresponding index value in the DLT, so an error occurs, and the non-effective pixel value cannot be reconstructed by the inverse transform, so the DC component and the original pixel are generated.
  • the probability of the same value directly affects the probability of error occurrence during the DLT transformation of a single image block, and also affects the total error in the DLT transformation process of the entire image.
  • the video encoding device may be used to perform encoding processing on an image to be processed (such as a depth map) to obtain encoded data, and the video decoding device may perform decoding processing on the encoded data of the image to be processed to complete image reconstruction.
  • an image to be processed such as a depth map
  • the video decoding device may perform decoding processing on the encoded data of the image to be processed to complete image reconstruction.
  • the video encoding device may be in a different device from the video decoding device, or may be in the same device.
  • the video encoding device may be the same device as the video decoding device, and the device can implement the encoding function of the video encoding device and the decoding function of the video decoding device.
  • an embodiment of the present invention provides an image processing method, which is applied to a multimedia field, and may be specifically used for encoding a depth image, and may be applied to video coding in an image processing system provided by the foregoing embodiment of the present invention.
  • the device includes the following steps:
  • the depth query table includes at least one index value and a pixel value corresponding to each index value, where the index value is used to represent a pixel value corresponding to the index value itself, and the depth query table may be preset;
  • the at least one image processing mode may include one or more of a planar mode, a direct current component mode, and an angular mode, the angular mode including 33 angular modes.
  • the at least one image processing mode may also include image processing modes other than the 35 image processing modes.
  • the at least one image processing mode may include 35 image processing modes of a planar mode, a direct current component mode, and 33 angular modes, or may include only 33 angular modes.
  • the coded depth image is encoded using each of the at least one image processing mode, and the image processing mode conforming to the preset rule is selected in conjunction with the subsequent steps.
  • the angle mode includes 33 angle modes.
  • the original image block is a square image block having a side length of 2N pixels, and N is a positive even number.
  • each image processing mode From the corresponding predicted image block.
  • the predicted image block is a square image block having a side length of 2N pixels, and N is a positive even number; and each of the image processing modes corresponding to the predicted image block is predetermined.
  • the predicted image blocks corresponding to each of the image processing modes are acquired according to the at least one image processing mode.
  • the reason for this is that the state of the intra-frame reference pixel filter function is on, and the higher the coding gain can be obtained when the state of the intra-frame reference pixel filter function is off.
  • the value of the reference pixel directly affects the performance of the present invention.
  • the closer the depth pixel value of the first mapping is to the original pixel value the smaller the error of the DLT transform.
  • the intra-frame reference pixel filter smoothes the value of the reference pixel, affecting the effect of using a single reference pixel value as a predictive DC (not conducive to reducing the error caused by the DLT transform), so when the intra-frame reference pixel filter is turned off, more High gain.
  • the predicted DC component may be obtained according to the pixel value of the pixel in the lower right corner of the predicted image block; or the predicted DC component may be obtained according to the pixel value of one pixel of the central region of the predicted image block, the predicted image
  • the block center area is a square image area having a side length of 2 pixels at the center of the predicted image block.
  • the pixel value of a single pixel is used as the predicted DC component, the large probability DLT transformation error caused by the averaging of the pixel values is avoided, and the transformation error generated by the DLT can be effectively reduced, thereby improving the reconstructed image block when generating the depth image. Coding performance.
  • the image processing mode is an angle mode, obtaining a predicted DC component according to a pixel value of one pixel in the predicted image block;
  • the predicted DC component is obtained according to a pixel value of one pixel in the predicted image block.
  • the predicted DC component is obtained according to a pixel value of one pixel in the predicted image block.
  • a further alternative of this step, relative to other schemes of this step, is to more effectively avoid the pixel value averaging because more image processing modes use pixel values of a single pixel as the predicted DC component.
  • the large probability DLT transformation error effectively reduces the transformation error generated by the DLT in the SDC image processing of the depth map, thereby improving the coding performance of the image processing result, and is the best alternative for this step.
  • an average value of pixel values of four corner pixel points of the predicted image block may be used as a predicted DC component.
  • This additional option is combined with the improvement of the prior art, and can also reduce the transformation error generated by the DLT in the SDC image processing of the depth map, thereby improving the coding performance of the image processing result, and the prior art
  • the combination facilitates the promotion of the present invention.
  • the range of depth pixel values is D 0 -D 7 , where D 0 , D 4 , D 7 are pixel values existing in the depth image; D 1 , D 2 , D 3 D 5 and D 6 are pixel values that do not exist in the original depth image.
  • the index values I 0 and I 1 corresponding to the pixel values D 0 , D 4 , and D 7 are included in the preset depth query table. I 2 .
  • the pixel value is obtained D 0, D 4, D 7 each corresponding to the index value I 0, I 1, I 2, as a DC prediction
  • the index value corresponding to the component it can also be referred to FIG. 4: if the predicted DC component is not D 0 , D 4 , D 7 and is D 1 , D 2 , D 3 , D 5 , D 6 , the neighboring pixel values are selected according to a predetermined rule.
  • the prediction index value is used as a prediction index value corresponding to the predicted DC component.
  • the original DC component can be obtained according to the pixel value of one pixel in the original image block.
  • the original DC component may be obtained according to the pixel value of the pixel in the lower right corner of the original image block; or the original DC component may be obtained according to the pixel value of one pixel of the central region of the original image block, the original
  • the image block central area is a square image area having a side length of 2 pixels at the center of the original image block.
  • the selection of the pixel points in the acquisition mode of the original DC component may refer to obtaining the position of the pixel point of the predicted DC component, or may not refer to the pixel if the original DC component reference is obtained. The point is the same as the pixel point for obtaining the predicted DC component reference, and the SDC has the best coding performance.
  • the original index value may be obtained by referring to the method for obtaining the predicted index value in the step 104, and the predicted DC component is replaced by the original DC component, and the predicted index value is replaced with the original index value, and other parts are the same. This will not be repeated here.
  • a rate distortion cost of the image processing mode corresponding to the predicted image block in the at least one image processing mode may be calculated according to the predicted image block; and then, according to the rate distortion cost, the at least one is An image processing mode in which the rate distortion cost is the smallest in the image processing mode is a preferred image processing mode.
  • the at least one image processing mode includes a total of 35 image processing modes of the planar mode, the direct current component mode, and the 33 angular modes. If calculated, the direct current component mode brings a minimum rate distortion cost. Then, the DC component mode is selected as the preferred image processing mode at this time.
  • the encoded data may also be sent to reconstruct a depth image according to the encoded data, where the encoded data includes the preferred image processing mode, an index residual value corresponding to the preferred image processing mode, and the depth Query list
  • the image processing method provided by the embodiment of the present invention can obtain an original image block of a depth image to be encoded, at least one image processing mode and a depth query table by using a video encoding device, acquire a predicted image block according to the image processing mode, and obtain the predicted image block according to the image processing mode.
  • the obtained information obtains the original DC component, and the predicted DC component is obtained according to the pixel value of one pixel in the predicted image block in combination with the specific image processing mode, and the index residual value is calculated according to the original DC component and the predicted DC component.
  • the probability that a single pixel value is a valid pixel value is much greater than the probability that multiple pixel values are effective pixels, so the SDC encoding process is reduced.
  • the transform error generated by the DLT in the middle improves the coding performance of the image processing result.
  • an embodiment of the present invention provides an image processing method, which is applied to a multimedia field, and may be specifically used for image processing of a depth image, and may be applied to a decoding end of an image processing system provided by the foregoing embodiment of the present invention. , including the following steps:
  • the depth query table includes at least one index value and a pixel value corresponding to each index value, where the index value is used to represent a pixel value corresponding to the index value itself, and the depth query table may be preset;
  • the image processing mode may include one of a planar mode, a direct current component mode, and an angular mode, and the angular mode includes 33 angular modes.
  • the image processing mode may also be an image processing mode other than the 35 image processing modes.
  • the angle mode includes 33 angle modes.
  • the original image block is a square image block having a side length of 2N pixels, and N is a positive even number.
  • the predicted image block is a square image block having a side length of 2N pixels, and N is a positive even number; and each of the image processing modes corresponding to the predicted image block is predetermined.
  • the predicted image block may be acquired according to the image processing mode.
  • the reason for this is that the state of the intra-frame reference pixel filter function is on, and the higher the coding gain can be obtained when the state of the intra-frame reference pixel filter function is off.
  • the value of the reference pixel directly affects the performance of the present invention.
  • the closer the depth pixel value of the first mapping is to the original pixel value the smaller the error of the DLT transform.
  • the intra-frame reference pixel filter smoothes the value of the reference pixel, affecting the effect of using a single reference pixel value as a predictive DC (not conducive to reducing the error caused by the DLT transform), so when the intra-frame reference pixel filter is turned off, more High gain.
  • the predicted DC component may be obtained according to the pixel value of the pixel in the lower right corner of the predicted image block; or the predicted DC component may be obtained according to the pixel value of one pixel of the central region of the predicted image block, the prediction
  • the image block central area is a square image area having a side length of 2 pixels at the center of the predicted image block.
  • the pixel value of a single pixel is used as the predicted DC component, the large probability DLT transformation error caused by the averaging of the pixel values is avoided, and the transformation error generated by the DLT can be effectively reduced, thereby improving the reconstructed image block when generating the depth image. Coding performance.
  • the image processing mode is an angle mode, obtaining a predicted DC component according to a pixel value of one pixel in the predicted image block;
  • the predicted DC component is obtained according to a pixel value of one pixel in the predicted image block.
  • the predicted DC component is obtained according to a pixel value of one pixel in the predicted image block.
  • a further alternative of this step, relative to other schemes of this step, is to more effectively avoid the pixel value averaging because more image processing modes use pixel values of a single pixel as the predicted DC component.
  • the large probability DLT transformation error effectively reduces the transformation error generated by the DLT in the SDC image processing of the depth map, thereby improving the coding performance of the image processing result, and is the best alternative for this step.
  • an average value of pixel values of four corner pixel points of the predicted image block may be used as a predicted DC component.
  • This additional option is combined with the improvement of the prior art, and can also reduce the transformation error generated by the DLT in the SDC image processing of the depth map, thereby improving the coding performance of the image processing result, and the prior art
  • the combination facilitates the promotion of the present invention.
  • the range of depth pixel values is D 0 -D 7 , where D 0 , D 4 , D 7 are pixel values existing in the depth image; D 1 , D 2 , D 3 D 5 and D 6 are pixel values that do not exist in the original depth image.
  • the index values I 0 and I 1 corresponding to the pixel values D 0 , D 4 , and D 7 are included in the preset depth query table. I 2 .
  • the pixel value is obtained D 0, D 4, D 7 each corresponding to the index value I 0, I 1, I 2, as a DC prediction
  • the index value corresponding to the component it can also be referred to FIG. 4: if the predicted DC component is not D 0 , D 4 , D 7 and is D 1 , D 2 , D 3 , D 5 , D 6 , the neighboring pixel values are selected according to a predetermined rule.
  • the prediction index value is used as a prediction index value corresponding to the predicted DC component.
  • the reconstructed DC component may be reversely acquired by referring to the depth query table and referring to the manner of obtaining the predicted index value in the step 504. That is, the DLT inverse transform is performed.
  • the double-headed arrow connection shown in FIG. 4 can be used.
  • the pixel value that can uniquely correspond to the index value in the depth query table can only be obtained as the reconstructed DC component according to the reconstruction index value.
  • the pixel value D 1 corresponding to the DC component is converted into the index value I 0 by DLT, and the index value I 0 can only be inversely transformed into the DC component corresponding to the pixel value D 0 .
  • the image processing method provided by the embodiment of the present invention can obtain the encoded data of the depth image to be reconstructed by using a video decoding device, where the encoded data includes an image processing mode, an index residual value, and a depth query table, and combined with a specific image processing mode. Obtaining a predicted DC component according to a pixel value of one pixel in the predicted image block, and then performing a DLT transform, and acquiring the reconstructed image block in combination with the index residual value and the DLT inverse transform to finally obtain the reconstructed image.
  • the probability that the value of the DC component is completely derived from the effective pixel is improved (When the average value of the pixel values of a plurality of pixels is used as the DC component, if the average value is a decimal number and the pixel value is not a decimal number, more errors are generated. Meanwhile, only the pixel values of all the reference pixel points are effective pixel values. When using the DC component derived from these pixels, the variation error caused by DLT is not generated.
  • the probability that a single pixel value is a valid pixel value is much larger than the probability that a plurality of pixel values are effective pixels, so the coding of the SDC is reduced.
  • the transformation error produced by the DLT in the process improves the coding performance of the image processing result.
  • an embodiment of the present invention provides an image processing method, which is applied to the field of multimedia, and can be specifically used for image processing of a depth image, and can be applied to the present invention.
  • the encoding end in the image processing system provided by the above embodiment includes the following steps:
  • the depth query table includes at least one index value and a pixel value corresponding to each index value, where the index value is used to represent a pixel value corresponding to the index value itself, and the depth query table may be preset;
  • the at least one image processing mode may include one or more of a planar mode, a direct current component mode, and an angular mode, the angular mode including 33 angular modes.
  • the at least one image processing mode may also include image processing modes other than the 35 image processing modes.
  • the at least one image processing mode may include 35 image processing modes of a planar mode, a direct current component mode, and 33 angular modes, or may include only 33 angular modes.
  • the coded depth image is encoded using each of the at least one image processing mode, and the image processing mode conforming to the preset rule is selected in conjunction with the subsequent steps.
  • the angle mode includes 33 angle modes.
  • the original image block is a square image with a side length of 2N pixels Block, N is a positive even number.
  • the intra reference pixel filter When the intra reference pixel filter is turned off, the predicted image blocks corresponding to each of the image processing modes are acquired according to the at least one image processing mode.
  • the predicted image block is a square image block having a side length of 2N pixels, and N is a positive even number; and each of the image processing modes corresponding to the predicted image block is predetermined.
  • the reason for this is that the state of the intra-frame reference pixel filter function is on, and the higher the coding gain can be obtained when the state of the intra-frame reference pixel filter function is off.
  • the value of the reference pixel directly affects the performance of the present invention.
  • the closer the depth pixel value of the first mapping is to the original pixel value the smaller the error of the DLT transform.
  • the intra-frame reference pixel filter smoothes the value of the reference pixel, affecting the effect of using a single reference pixel value as a predictive DC (not conducive to reducing the error caused by the DLT transform), so when the intra-frame reference pixel filter is turned off, more High gain.
  • the predicted DC component may be obtained according to a pixel value of one pixel of the central region of the predicted image block, where the predicted image block center region is a square with a side length of 2 pixels at the center of the predicted image block. Image area.
  • P (W-1)>>1, (W-1)>>1 is a pixel value of a pixel point of an upper left corner of the central region of the predicted image block, and W is a value of a side length of the predicted image block;
  • P W>>1, W>>1 is the predicted image.
  • the pixel value of the pixel in the lower right corner of the block center area, and W is the side length value of the predicted image block.
  • the pixel value of a single pixel is used as the predicted DC component, the large probability DLT transformation error caused by the averaging of the pixel values is avoided, and the transformation error generated by the DLT can be effectively reduced, thereby improving the reconstructed image block when generating the depth image. Coding performance.
  • an average value of pixel values of four corner pixel points of the predicted image block may be used as a predicted DC component.
  • the image processing mode is the planar mode or the DC component mode
  • the pixel value, >>2 indicates the displacement operation in computer calculation, which is 2 bits shifted to the right. Because the computer calculates the commonly used binary calculation method, shifting 2 bits to the right is equivalent to dividing Pred DC by 2 times. Square, that is, divided by 4, at this time, because 4 pixel points are selected, such calculation is equivalent to averaging the pixel values of 4 pixel points.
  • This additional option is combined with the improvement of the prior art, and can also reduce the transformation error generated by the DLT in the SDC image processing of the depth map, thereby improving the coding performance of the image processing result, and the prior art
  • the combination facilitates the promotion of the present invention.
  • the center pixel can have 16-17 reference pixels in different positions, and the predicted pixel values of the predicted values of these positions have a wide range of positions.
  • the preferred mode can be selected when the VSO performs the RDO selection (eg, the selection of each image processing mode based on the rate distortion cost).
  • the probability that a single pixel value exists in the original depth image is much more likely to exist in the original depth image, so the error caused by the DLT transform and the inverse transform is also much smaller.
  • the range of depth pixel values is D 0 -D 7 , where D 0 , D 4 , D 7 are pixel values existing in the depth image; D 1 , D 2 , D 3 D 5 and D 6 are pixel values that do not exist in the original depth image.
  • the index values I 0 and I 1 corresponding to the pixel values D 0 , D 4 , and D 7 are included in the preset depth query table. I 2 .
  • the pixel value is obtained D 0, D 4, D 7 each corresponding to the index value I 0, I 1, I 2, as a DC prediction
  • the index value corresponding to the component it can also be referred to FIG. 4: if the predicted DC component is not D 0 , D 4 , D 7 and is D 1 , D 2 , D 3 , D 5 , D 6 , the neighboring pixel values are selected according to a predetermined rule.
  • the prediction index value is used as a prediction index value corresponding to the predicted DC component.
  • OrigDC is the original DC component
  • P W-1, W-1 are in the original image block.
  • the pixel value of the pixel in the lower right corner, and W is the value of the side length of the original image block.
  • the step may further obtain an original DC component according to a pixel value of a pixel of the central region of the original image block, where the original image block central region is A square image area having a side length of 2 pixels at the center of the original image block.
  • OrigDC is the original DC component
  • P W>>1 is the original image.
  • W is the value of the side length of the original image block.
  • the selection of the pixel points in the acquisition mode of the original DC component may refer to obtaining the position of the pixel point of the predicted DC component, or may not refer to the pixel if the original DC component reference is obtained. The point is the same as the pixel point for obtaining the predicted DC component reference, and the SDC has the best coding performance.
  • the original image block size is 64 ⁇ 64 or 32 ⁇ 32 (the unit is a pixel)
  • the original block is first downsampled by 2:1, and then the average value of all pixel values of the downsampled block is used as the original DC.
  • Component ie If the original image block size is smaller than 32 ⁇ 32, the mean value of all pixel values of the original block is directly used as the original DC component, that is, This alternative can be used as a default alternative, image processing is not preferred, but most commonly used, the changes to the prior art are minimal, and the most popular.
  • the original index value may be obtained by referring to the method for obtaining the predicted index value in step 604, and the predicted DC component is replaced by the original DC component, and the predicted index value is replaced with the original index value, and the other parts are the same, and no longer Narration.
  • a rate distortion cost of the image processing mode corresponding to the predicted image block in the at least one image processing mode may be calculated according to the predicted image block; and then, according to the rate distortion cost, the at least one is An image processing mode in which the rate distortion cost is the smallest in the image processing mode is a preferred image processing mode.
  • the at least one image processing mode includes a total of 35 image processing modes of the planar mode, the direct current component mode, and the 33 angular modes. If calculated, the direct current component mode brings a minimum rate distortion cost. Then, the DC component mode is selected as the preferred image processing mode at this time.
  • the encoded data may also be sent to reconstruct a depth image according to the encoded data, where the encoded data includes the preferred image processing mode, an index residual value corresponding to the preferred image processing mode, and the depth Query the table.
  • the image processing method provided by the embodiment of the present invention can obtain an original image block of a depth image to be encoded, at least one image processing mode and a depth query table by using a video encoding device, and acquire a predicted image block according to an image processing mode, and combine the specific image.
  • the processing mode acquires a predicted direct current component according to a pixel value of a pixel point in a lower right corner of the predicted image block, and extracts an original direct current component according to a pixel value of a pixel point in a lower right corner of the original image block, and performs an index on the DLT according to the original direct current component and the predicted direct current component.
  • the residual value is used to reconstruct the encoded depth image from the index residual value.
  • the conversion error generated by the DLT in the SDC encoding process is reduced, and the RDO can be selected in the VSO. Selecting a more appropriate encoding result improves the encoding performance of the image processing result.
  • an embodiment of the present invention provides an image processing method, which is applied to the field of multimedia, and can be specifically used for image processing of a depth image, which can be applied to the present invention.
  • the decoding end in the image processing system provided by the above embodiment includes the following steps:
  • Acquire encoded data of a depth image to be reconstructed where the encoded data includes an image processing mode, an index residual value, and a depth lookup table.
  • the depth query table includes at least one index value and a pixel value corresponding to each index value, where the index value is used to represent a pixel value corresponding to the index value itself, and the depth query table may be preset;
  • the image processing mode may include one of a planar mode, a direct current component mode, and an angular mode, and the angular mode includes 33 angular modes.
  • the image processing mode may also be an image processing mode other than the 35 image processing modes.
  • the angle mode includes 33 angle modes.
  • the original image block is a square image block having a side length of 2N pixels, and N is a positive even number.
  • the predicted image block is a square image block having a side length of 2N pixels, and N is a positive even number; and each of the image processing modes corresponding to the predicted image block is predetermined.
  • the reason for this is that the state of the intra-frame reference pixel filter function is on, and the higher the coding gain can be obtained when the state of the intra-frame reference pixel filter function is off.
  • the value of the reference pixel directly affects the performance of the present invention.
  • the closer the depth pixel value of the first mapping is to the original pixel value the smaller the error of the DLT transform.
  • the intra-frame reference pixel filter smoothes the value of the reference pixel, affecting the effect of using a single reference pixel value as a predictive DC (not conducive to reducing the error caused by the DLT transform), so when the intra-frame reference pixel filter is turned off, more High gain.
  • the predicted DC component may be obtained according to a pixel value of one pixel of the central region of the predicted image block, where the predicted image block center region is 2 pixels at the center of the predicted image block. Square image area.
  • P (W-1)>>1, (W-1)>>1 is a pixel value of a pixel point of an upper left corner of the central region of the predicted image block, and W is a value of a side length of the predicted image block;
  • P W>>1, W>>1 is the predicted image.
  • the pixel value of the pixel in the lower right corner of the block center area, and W is the side length value of the predicted image block.
  • the pixel value of a single pixel is used as the predicted DC component, the large probability DLT transformation error caused by the averaging of the pixel values is avoided, and the transformation error generated by the DLT can be effectively reduced, thereby improving the reconstructed image block when generating the depth image. Coding performance.
  • an average value of pixel values of four corner pixel points of the predicted image block may be used as a predicted DC component.
  • the image processing mode is the planar mode or the DC component mode
  • the pixel value, >>2 indicates the displacement operation in computer calculation, which is 2 bits shifted to the right. Because the computer calculates the commonly used binary calculation method, shifting 2 bits to the right is equivalent to dividing Pred DC by 2 times. Square, that is, divided by 4, at this time, because 4 pixels are selected, such calculation is equivalent to averaging the pixel values of 4 pixels.
  • This additional option is combined with the improvement of the prior art, and can also reduce the transformation error generated by the DLT in the SDC image processing of the depth map, thereby improving the coding performance of the image processing result, and the prior art
  • the combination facilitates the promotion of the present invention.
  • the pixel values in the lower right corner can have 33 different reference pixels in 33 angular modes, and the center pixel can have 16-17 reference pixels in different positions.
  • the reference pixel value of the value has a wide range of positions, and the VDO (View Synthesis Optimization) can select the preferred mode when performing RDO (Rate Distortion Optimization) selection (for example, according to the rate distortion cost) Selection of each image processing mode).
  • RDO Rate Distortion Optimization
  • the probability that a single pixel value exists in the original depth image is much more likely to exist in the original depth image, so the error caused by the DLT transform and the inverse transform is also much smaller.
  • the range of depth pixel values is D 0 -D 7 , where D 0 , D 4 , D 7 are pixel values existing in the depth image; D 1 , D 2 , D 3 D 5 and D 6 are pixel values that do not exist in the original depth image.
  • the index values I 0 and I 1 corresponding to the pixel values D 0 , D 4 , and D 7 are included in the preset depth query table. I 2 .
  • the pixel value is obtained D 0, D 4, D 7 each corresponding to the index value I 0, I 1, I 2, as a DC prediction
  • the index value corresponding to the component it can also be referred to FIG. 4: if the predicted DC component is not D 0 , D 4 , D 7 and is D 1 , D 2 , D 3 , D 5 , D 6 , the neighboring pixel values are selected according to a predetermined rule.
  • the prediction index value is used as a prediction index value corresponding to the predicted DC component.
  • the reconstructed DC component may be reversely acquired by referring to the depth query table and referring to the manner of obtaining the predicted index value in the step 804. That is, the DLT inverse transform is performed.
  • the double-headed arrow connection shown in FIG. 4 can be used.
  • the pixel value that can uniquely correspond to the index value in the depth query table can only be obtained as the reconstructed DC component according to the reconstruction index value.
  • the pixel value D 1 corresponding to the DC component is converted into the index value I 0 by DLT, and the index value I 0 can only be inversely transformed into the DC component corresponding to the pixel value D 0 .
  • the image processing method provided by the embodiment of the present invention can obtain the encoded data of the depth image to be reconstructed by using a video decoding device, where the encoded data includes an image processing mode, an index residual value, and a depth query table, and combined with a specific image processing mode. Obtaining a predicted DC component according to a pixel value of one pixel in the predicted image block, and then performing a DLT transform, and acquiring the reconstructed image block in combination with the index residual value and the DLT inverse transform to finally obtain the reconstructed image.
  • the conversion error generated by the DLT in the SDC encoding process is reduced, and the RDO can be selected in the VSO. Selecting a more appropriate encoding result improves the encoding performance of the image processing result.
  • Table 1.1, Table 1.2, Table 2.1, and Table 2.2 are provided.
  • Tables 3 and 4 are used to illustrate the beneficial effects of the various alternatives in step 803.
  • the left column of the table is the name of the test video used for testing.
  • Tables 1.1 and 1.2 show the test results of the invention using only AI (All Intra, all intra mode) 2 frames in the 33 angle modes of the depth map.
  • Table 1.1 shows the scheme using the prediction block. The pixel value in the lower left corner of the center block is used as the prediction DC, and the average coding gain is 0.10%.
  • the scheme shown in Table 1.2 uses the pixel value of the lower right corner of the prediction block as the prediction DC, resulting in an average coding gain of 0.08%.
  • the present invention is most effective for the angle mode, but since the single pixel value prediction reduces the error of the DLT transformation, on the basis of the application scenarios of Table 1.1 and Table 1.2, the application mode of the present invention is further expanded to the DC mode and the planar mode. Come to a certain gain.
  • Table 2.1 and Table 2.2 show the test results under the AI 2 frame. Table 2.1 shows that the DC mode and the angle mode of the depth map use the lower left corner pixel in the center area for predicting DC, and the planar mode uses four corner pixel mean values for predicting DC.
  • the average coding gain is 0.15%;
  • Table 2.2 shows that the planar mode and the angle mode of the depth map use the DC of the lower right corner of the prediction block for prediction, and the DC mode uses the average of four angular pixels for the prediction DC.
  • the average coding gain is 0.11%.
  • Table 4 shows the experimental results under the setting conditions of Table 3 when the intra-frame reference pixel filter function is turned off, and the average coding gain is 0.31%.
  • the technical solution shown in Table 4 is the present invention. Preferred technical solution.
  • the above tables can display color video of only one viewpoint when encoding two-dimensional video; three-dimensional video needs to separately encode color and depth video of multiple viewpoints, and the video of different viewpoints can make the viewer feel stereo.
  • Video from different viewpoints is equivalent to the same scene seen from different directions.
  • the depth video reflects the distance information of the scene, which is not used for direct viewing, but is used to assist the color video to synthesize other viewpoint videos. Therefore, the encoding quality of the depth image is not directly checked, but the quality of the depth image and the color video is evaluated by the quality of the synthesized viewpoint.
  • the color and depth video of three viewpoints are simultaneously encoded, and the video of any viewpoint is synthesized by the depth video and the color video, and the overall coding performance is measured by the quality of the synthesized video.
  • the video0, video1, and video2 in Tables 1.1 and 1.2 respectively represent color video of three different viewpoints, and the corresponding experimental data represents the reconstruction process of the reconstructed image composed of the reconstructed image obtained by encoding the original image in the original video.
  • PSNR Peak Signal to Noise Ratio
  • Video only represents color video, and total represents color plus depth video.
  • Video PSNR/video bitrate represents the ratio of the PSNR of the color video to the number of bits of the color video;
  • video PSNR/total bitrate represents the ratio of the PSNR of the color video to the sum of the bits of the color video and the depth video;
  • synth PSNR/total bitrate represents the composite video The PSNR and the ratio of the number of bits of the color video to the depth video, this value best reflects the performance of the encoding.
  • the prediction in the image processing method provided by the embodiment of the present invention, in the decoding process of the encoded data by the video decoding device, for any of the image processing modes, the prediction can be obtained according to the pixel value of one pixel in the predicted image block.
  • a DC component, and the predicted DC component used by the decoding end is specifically acquired in the same manner as the encoding end, so that the transform error generated by the DLT in the decoding process can be reduced, thereby improving the encoding when reconstructing the image block of the depth image. performance.
  • an embodiment of the present invention provides a video encoding apparatus, which is applicable to the field of multimedia, and can be used for image processing of a depth image, and can be applied to an image processing system provided by an embodiment of the present invention.
  • the above-described embodiments of the invention are used for an image processing method provided by a video encoding apparatus in an image processing system, and include the following structures:
  • An obtaining unit 901 configured to acquire an original image block of a depth image to be encoded, at least one An image processing mode and a depth query table, wherein the depth query table includes at least one index value and a pixel value corresponding to each index value, the index value being used to represent a pixel value corresponding to the index value itself;
  • the obtaining unit 901 is further configured to acquire, according to the at least one image processing mode acquired by the acquiring unit 901, a predicted image block corresponding to each image processing mode.
  • the predicted DC component unit 902 is configured to obtain a predicted DC component according to a pixel value of one pixel in the predicted image block acquired by the acquiring unit 901;
  • the prediction index value unit 903 is configured to obtain a prediction index value according to the depth query table acquired by the acquiring unit 901 and the predicted DC component acquired by the DC component unit.
  • the original DC component unit 904 is configured to acquire an original DC component according to the original image block acquired by the acquiring unit 901;
  • the original index value unit 905 is configured to obtain an original index value according to the depth query table acquired by the acquiring unit 901 and the original DC component acquired by the original DC component unit 904;
  • the index residual value unit 906 is configured to obtain the index residual value by subtracting the original index value obtained by the original index value unit 905 from the predicted index value obtained by the prediction index value unit 903;
  • the selecting unit 907 is configured to select a preferred image processing mode from the at least one image processing mode according to the predicted image block and the preset rule acquired by the acquiring unit 901;
  • a coding unit 908 the preferred image processing mode selected by the preferred unit 907, an index residual value corresponding to the preferred image processing mode acquired by the index residual value unit 906, and the acquiring unit
  • the depth lookup table obtained by 901 is encoded to obtain encoded data.
  • the video encoding device may also transmit the encoded data outward such that other devices generate a reconstructed depth image from the encoded data.
  • the predicted DC component unit 902 is specifically configured to obtain a predicted DC component according to a pixel value of a pixel in a lower right corner of the predicted image block acquired by the acquiring unit 901;
  • the predicted DC component unit 902 is specifically configured to acquire a predicted DC component according to a pixel value of one pixel of the central region of the predicted image block acquired by the acquiring unit 901, where the predicted image block center region is the prediction A square image area with a side length of 2 pixels at the center of the image block.
  • the predicted DC component unit 902 is further configured to: when the image processing mode acquired by the acquiring unit 901 is an angle mode, according to a pixel point in the predicted image block acquired by the acquiring unit 901 The pixel value gets the predicted DC component.
  • the predicted DC component unit 902 is further configured to be acquired according to the acquiring unit 901 when the image processing mode acquired by the acquiring unit 901 is a planar mode.
  • the pixel value of one pixel in the predicted image block acquires a predicted DC component.
  • the predicted DC component unit 902 is further configured to: when the image processing mode acquired by the acquiring unit 901 is a DC component mode, according to the pixel in the predicted image block acquired by the acquiring unit 901 The pixel value of the point acquires the predicted DC component.
  • the original DC component unit 904 is specifically configured to acquire an original DC component according to a pixel value of one pixel in the original image block acquired by the acquiring unit 901.
  • the original DC component unit 904 is specifically configured to acquire an original DC component according to a pixel value of a pixel in a lower right corner of the original image block acquired by the acquiring unit 901;
  • the original DC component unit 904 is specifically configured to acquire an original DC component according to a pixel value of a pixel of the central region of the original image block acquired by the acquiring unit 901, where the original image block central region is the original A square image area with a side length of 2 pixels at the center of the image block.
  • the acquiring unit 901 is configured to acquire, according to the at least one image processing mode acquired by the acquiring unit 901, a predicted image corresponding to each image processing mode when the intra reference pixel filter is turned off. Piece.
  • the preferred unit 907 specifically includes:
  • a rate-distortion sub-unit 9071 configured to calculate a rate-distortion cost of an image processing mode corresponding to the predicted image block in the at least one image processing mode according to the predicted image block acquired by the acquiring unit 901;
  • the subunit 9072 is configured to select, according to the rate distortion cost calculated by the rate distortion subunit 9071, an image processing mode in which the rate distortion cost of the at least one image processing mode acquired by the acquiring unit 901 is the smallest. Image processing mode.
  • the video encoding apparatus can acquire an original image block of a depth image to be encoded, at least one image processing mode, and a depth query table, and acquire a predicted image block according to the image processing mode, according to the specific image processing mode. Predicting the pixel value of the pixel in the lower right corner of the image block to obtain the predicted DC component, and extracting the original DC component according to the pixel value of the pixel pixel in the lower right corner of the original image block, and calculating the index residual value according to the original DC component and the predicted DC component, so that the index residual value is The encoded depth image is reconstructed based on the index residual value.
  • the conversion error generated by the DLT in the SDC encoding process is reduced, and can be selected when the VSO performs the RDO selection.
  • Appropriate coding results which improve the coding performance of image processing results.
  • an embodiment of the present invention provides a video decoding device, which is applicable to the field of multimedia, and can be used for image processing of a depth image, and can be applied to an image processing system provided by an embodiment of the present invention.
  • the above-described embodiments of the invention are used for an image processing method provided by a video decoding device in an image processing system, and include the following structures:
  • An acquiring unit 1101 configured to acquire encoded data of a depth image to be reconstructed, where the encoding
  • the data includes an image processing mode, an index residual value, and a depth query table, where the depth query table includes at least one index value and a pixel value corresponding to each index value, the index value is used to indicate that the index value itself corresponds to Pixel value;
  • the obtaining unit 1101 is further configured to acquire a predicted image block according to the image processing mode acquired by the acquiring unit 1101;
  • a prediction DC component unit 1102 configured to acquire a predicted DC component according to a pixel value of one pixel in the predicted image block acquired by the acquiring unit 1101;
  • the prediction index value unit 1103 is configured to obtain a prediction index value according to the depth query table acquired by the acquiring unit 1101 and the predicted DC component acquired by the DC component unit.
  • the index value unit 1104 is configured to add the index residual value obtained by the acquiring unit 1101 to the prediction index value obtained by the prediction index value unit 1103 to obtain a reconstruction index value;
  • the reconstructed DC component unit 1105 is configured to obtain a reconstructed pixel value according to the depth query table obtained by the acquiring unit 1101 and the reconstructed index value obtained by the reconstructed index value unit 1104, and use the reconstructed pixel value as a reconstructed DC Component
  • a depth residual difference unit 1106, configured to subtract the reconstructed direct current component obtained by the reconstructed direct current component unit 1105 from the predicted direct current component to obtain a depth residual value
  • the reconstruction unit 1107 is configured to obtain, by using the pixel value of each pixel of the predicted image block acquired by the acquiring unit 1101, the depth residual value obtained by the depth residual value unit 1106 to obtain a reconstructed image block, The reconstructed image block is used to reconstruct the depth image to be reconstructed.
  • the predicted DC component unit 1102 is specifically configured to acquire a predicted DC component according to a pixel value of a pixel in a lower right corner of the predicted image block acquired by the acquiring unit 1101;
  • the predicted DC component unit 1102 is specifically configured to acquire a predicted DC component according to a pixel value of one pixel of the central region of the predicted image block acquired by the acquiring unit 1101, where the predicted image block center region is the prediction A square image area with a side length of 2 pixels at the center of the image block.
  • the predicted DC component unit 1102 is further configured to: when the image processing mode acquired by the acquiring unit 1101 is an angle mode, according to a pixel point in the predicted image block acquired by the acquiring unit 1101 The pixel value gets the predicted DC component.
  • the predicted DC component unit 1102 is further configured to: when the image processing mode acquired by the acquiring unit 1101 is a planar mode, according to the pixel in the predicted image block acquired by the acquiring unit 1101 The pixel value gets the predicted DC component.
  • the predicted DC component unit 1102 is further configured to: when the image processing mode acquired by the acquiring unit 1101 is a DC component mode, according to the pixel in the predicted image block acquired by the acquiring unit 1101 The pixel value of the point acquires the predicted DC component.
  • the acquiring unit 1101 is specifically configured to acquire a predicted image block according to the image processing mode acquired by the acquiring unit 1101 when the intra reference pixel filter is turned off.
  • the video decoding apparatus can obtain the encoded data of the depth image to be reconstructed, where the encoded data includes an image processing mode, an index residual value, and a depth query table, and is combined with the specific image processing mode according to the predicted image.
  • the pixel value of one pixel in the block acquires the predicted DC component, and then performs DLT transformation, and combines the index residual value and the DLT inverse transform to obtain the reconstructed image block, so as to finally obtain the reconstructed image.
  • the conversion error generated by the DLT in the SDC encoding process is reduced, and the RDO can be selected in the VSO. Selecting a more appropriate encoding result improves the encoding performance of the image processing result.
  • An embodiment of the present invention provides a video encoding apparatus, which is applicable to the field of multimedia, and can be used for image processing of a depth image, and can be applied to an image processing system provided by an embodiment of the present invention, which can be combined with the above embodiment of the present invention.
  • the video encoding device can be embedded or itself a micro-processing computer, such as a general-purpose computer, a customized computer, a mobile phone terminal or a tablet computer.
  • the video encoding apparatus 12001 includes at least one data port 12011, a processor 12012, a memory 12013, and a bus 12014. The at least one data port 12011, the processor 12012, and the memory 12013 are connected by the bus 12014 and complete communication with each other.
  • the bus 12014 may be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component) bus, or an EISA (Extended Industry Standard Architecture) bus.
  • the bus 12014 can be divided into an address bus, a data bus, a control bus, and the like. For ease of representation, only one thick line is shown in Figure 12, but it does not mean that there is only one bus or one type of bus. among them:
  • the memory 12013 is for storing executable program code including computer operating instructions.
  • the memory 12013 may include a high speed RAM memory and may also include a non-volatile memory such as at least one disk memory.
  • the processor 12012 may be a central processing unit (CPU), or an application specific integrated circuit (ASIC), or one or more configured to implement the embodiments of the present invention. integrated circuit.
  • CPU central processing unit
  • ASIC application specific integrated circuit
  • the data port 12011 is configured to acquire an original image block of a depth image to be encoded, at least one image processing mode, and a depth query table, where the depth query table includes at least one index value and corresponding to each index value. a pixel value, where the index value is used to represent a pixel value corresponding to the index value itself;
  • the data port 12011 is further configured to acquire, according to the at least one image processing mode acquired by the data port 12011, a predicted image block corresponding to each image processing mode.
  • the processor 12012 is configured to obtain a predicted DC component according to a pixel value of one pixel in the predicted image block acquired by the data port 12011;
  • the processor 12012 is configured to obtain a prediction index value according to the depth query table acquired by the data port 12011 and the predicted DC component acquired by the DC component unit.
  • the processor 12012 is configured to acquire an original DC component according to the original image block acquired by the data port 12011;
  • the processor 12012 is configured to obtain an original index value according to the depth query table obtained by the data port 12011 and the original DC component acquired by the processor 12012;
  • the processor 12012 is configured to obtain the index residual value by subtracting the original index value acquired by the processor 12012 from the predicted index value obtained by the processor 12012;
  • the processor 12012 is configured to select a preferred image processing mode from the at least one image processing mode according to the predicted image block and a preset rule acquired by the data port 12011;
  • the processor 12012 is configured to select the preferred image processing mode selected by the processor 12012, an index residual value corresponding to the preferred image processing mode acquired by the processor 12012, and the data port 12011
  • the obtained depth lookup table is encoded to obtain encoded data.
  • the video encoding device may also transmit the encoded data outward such that other devices generate a reconstructed depth image from the encoded data.
  • the function of transmitting the encoded data may be performed by the data port 12011.
  • the processor 12012 is specifically configured to obtain a predicted DC component according to a pixel value of a pixel in a lower right corner of the predicted image block acquired by the data port 12011;
  • the processor 12012 is specifically configured to acquire a predicted DC component according to a pixel value of one pixel of the central region of the predicted image block acquired by the data port 12011, where the predicted image block central region is the predicted image block.
  • the processor 12012 is further configured to: when the image processing mode acquired by the data port 12011 is an angle mode, a pixel of one pixel in the predicted image block acquired according to the data port 12011 The value gets the predicted DC component.
  • the processor 12012 is further configured to: when the image processing mode acquired by the data port 12011 is a planar mode, a pixel of one pixel in the predicted image block acquired according to the data port 12011 The value gets the predicted DC component.
  • the processor 12012 is further configured to acquire, according to the data port 12011, when the image processing mode acquired by the data port 12011 is a DC component mode.
  • the pixel value of one pixel in the predicted image block acquires a predicted DC component.
  • the processor 12012 is specifically configured to obtain an original DC component according to a pixel value of one pixel in the original image block acquired by the data port 12011.
  • the processor 12012 is specifically configured to obtain an original DC component according to a pixel value of a pixel in a lower right corner of the original image block acquired by the data port 12011;
  • the processor 12012 is specifically configured to acquire an original DC component according to a pixel value of one pixel of the central region of the original image block acquired by the data port 12011, where the original image block central region is the original image block.
  • the data port 12011 is specifically configured to acquire, according to the at least one image processing mode acquired by the data port 12011, a predicted image corresponding to each image processing mode when the intra reference pixel filter is turned off. Piece.
  • the processor 12012 is configured to calculate, according to the predicted image block acquired by the data port 12011, a rate distortion of an image processing mode corresponding to the predicted image block in the at least one image processing mode. cost;
  • the processor 12012 is configured to select, according to the rate distortion cost calculated by the processor 12012, an image processing mode in which the rate distortion cost of the at least one image processing mode acquired by the data port 12011 is the smallest. Image processing mode.
  • the video encoding apparatus can acquire an original image block of a depth image to be encoded, at least one image processing mode, and a depth query table, and acquire a predicted image block according to the image processing mode, according to the specific image processing mode. Predicting the pixel value of the pixel in the lower right corner of the image block to obtain the predicted DC component, and extracting the original DC component according to the pixel value of the pixel pixel in the lower right corner of the original image block, and calculating the index residual value according to the original DC component and the predicted DC component, so that the index residual value is The encoded depth image is reconstructed based on the index residual value.
  • the conversion error generated by the DLT in the SDC encoding process is reduced, and the RDO can be selected in the VSO. Selecting a more appropriate encoding result improves the encoding performance of the image processing result.
  • the embodiment of the present invention provides a video decoding device, which is applicable to the field of multimedia, and can be used for image processing of a depth image, and can be applied to an image processing system provided by an embodiment of the present invention, which can be combined with the above embodiment of the present invention.
  • the video decoding device may be embedded or itself a microprocessor computer, such as a general-purpose computer, a custom machine, a mobile terminal, or a tablet.
  • the video decoding device 13001 includes at least one data port 13011, a processor 13012, a memory 13013, and a bus 13014.
  • the at least one data port 13011, the processor 13012, and the memory 13013 are connected by a bus 13014 and complete communication with each other.
  • the bus 13014 can be an ISA (Industry Standard Architecture) Industry standard architecture) bus, PCI (Peripheral Component) bus or EISA (Extended Industry Standard Architecture) bus.
  • the bus 13014 can be divided into an address bus, a data bus, a control bus, and the like. For ease of representation, only one thick line is shown in FIG. 13, but it does not mean that there is only one bus or one type of bus. among them:
  • the memory 13013 is for storing executable program code, the program code including computer operating instructions.
  • the memory 13013 may include a high speed RAM memory and may also include a non-volatile memory such as at least one disk memory.
  • the processor 13012 may be a central processing unit (CPU), or an application specific integrated circuit (ASIC), or one or more configured to implement the embodiments of the present invention. integrated circuit.
  • CPU central processing unit
  • ASIC application specific integrated circuit
  • the data port 13011 is configured to acquire encoded data of a depth image to be reconstructed, where the encoded data includes an image processing mode, an index residual value, and a depth query table, where the depth query table includes at least one index value and a pixel value corresponding to each index value, wherein the index value is used to represent a pixel value corresponding to the index value itself;
  • the data port 13011 is further configured to acquire a predicted image block according to the image processing mode acquired by the data port 13011;
  • the processor 13012 is configured to obtain a predicted DC component according to a pixel value of one pixel in the predicted image block acquired by the data port 13011;
  • the processor 13012 is configured to obtain a prediction index value according to the depth query table acquired by the data port 13011 and the predicted DC component acquired by the DC component unit.
  • the processor 13012 is configured to add the index residual value obtained by the data port 13011 to the predicted index value acquired by the processor 13012 to obtain a reconstruction index value;
  • the processor 13012 is configured to obtain a reconstructed pixel value according to the depth query table obtained by the data port 13011 and the reconstructed index value obtained by the processor 13012, and use the reconstructed pixel value as a reconstructed DC component;
  • the processor 13012 is configured to subtract the reconstructed DC component obtained by the processor 13012 from the predicted DC component to obtain a depth residual value
  • the processor 13012 is configured to obtain, by using the pixel value of each pixel of the predicted image block acquired by the data port 13011, the depth residual value obtained by the processor 13012 to obtain a reconstructed image block.
  • the reconstructed image block is used to reconstruct the depth image to be reconstructed.
  • the processor 13012 is specifically configured to obtain a predicted DC component according to a pixel value of a pixel in a lower right corner of the predicted image block acquired by the data port 13011;
  • the processor 13012 is specifically configured to acquire a predicted DC component according to a pixel value of one pixel of the central region of the predicted image block acquired by the data port 13011, where the predicted image block central region is the predicted image block.
  • the center is 2 pixels long Square image area.
  • the processor 13012 is further configured to: when the image processing mode acquired by the data port 13011 is an angle mode, the pixel of one pixel in the predicted image block acquired according to the data port 13011 The value gets the predicted DC component.
  • the processor 13012 is further configured to: when the image processing mode acquired by the data port 13011 is a planar mode, the pixel of one pixel in the predicted image block acquired according to the data port 13011 The value gets the predicted DC component.
  • the processor 13012 is further configured to: when the image processing mode acquired by the data port 13011 is a DC component mode, a pixel point in the predicted image block acquired according to the data port 13011 The pixel value obtains the predicted DC component.
  • the data port 13011 is specifically configured to acquire a predicted image block according to the image processing mode acquired by the data port 13011 when the intra reference pixel filter is turned off.
  • the video decoding apparatus can obtain the encoded data of the depth image to be reconstructed, where the encoded data includes an image processing mode, an index residual value, and a depth query table, and is combined with the specific image processing mode according to the predicted image.
  • the pixel value of one pixel in the block acquires the predicted DC component, and then performs DLT transformation, and combines the index residual value and the DLT inverse transform to obtain the reconstructed image block, so as to finally obtain the reconstructed image.
  • the conversion error generated by the DLT in the SDC encoding process is reduced, and the RDO can be selected in the VSO. Selecting a more appropriate encoding result improves the encoding performance of the image processing result.
  • Computer readable media includes both computer storage media and communication media including any medium that facilitates transfer of a computer program from one location to another.
  • a storage medium may be any available media that can be accessed by a computer.
  • the computer readable medium may include a RAM (Random Access Memory), a ROM (Read Only Memory), and an EEPROM (Electrically Erasable Programmable Read Only Memory).
  • any connection can suitably be a computer readable medium.
  • the software is transmitted from a website, server, or other remote source using coaxial cable, fiber optic cable, twisted pair, DSL (Digital Subscriber Line), or wireless technologies such as infrared, radio, and microwave, Then coaxial cable, fiber optic cable, twisted pair, DSL or such as infrared, wireless and Wireless technologies such as microwaves are included in the fixing of the associated medium.
  • the disc and the disc include a CD (Compact Disc), a laser disc, a compact disc, a DVD disc (Digital Versatile Disc), a floppy disc, and a Blu-ray disc, wherein the disc is usually magnetically copied,
  • the disc uses a laser to optically replicate the data. Combinations of the above should also be included within the scope of the computer readable media.
  • the image processing method, device and system provided by the embodiments of the present invention can be extended to other technologies related to DLT transformation of depth images, such as DLT-based DMM (Depth Modelling Mode) mode original DC.
  • DLT-based DMM Depth Modelling Mode

Landscapes

  • Engineering & Computer Science (AREA)
  • Multimedia (AREA)
  • Signal Processing (AREA)
  • Compression Or Coding Systems Of Tv Signals (AREA)

Abstract

提供一种图像处理方法、装置及系统，涉及多媒体领域，能减少对深度图进行SDC图像处理时DLT产生的变换误差。具体方案为：获取待编码深度图像的原始图像块、至少一种图像处理模式和深度查询表，获取每种图像处理模式各自对应的预测图像块；根据原始图像块获取原始直流分量；根据预测图像块中一个像素点的像素值获取预测直流分量；根据深度查询表和预测直流分量、原始直流分量获得索引残差值；根据预设规则从至少一种图像处理模式中选取优选的图像处理模式；对所述优选的图像处理模式、所述优选的图像处理模式对应的索引残差值和所述深度查询表进行编码以得到编码数据。

Description

一种图像处理方法、装置及系统 技术领域
本发明涉及多媒体领域，尤其涉及一种图像处理方法、装置及系统。
背景技术
SDC(Segment-wise Direct Component Coding，分区域的直流分量编码)是应用在3D-HEVC(High Efficiency Video Coding，高效视频编码标准)中的一种残差数据编码方法。使用SDC对深度图像进行图像处理时，通过建立DLT(Depth Lookup Table，深度查询表)编码索引残差值到编码数据中发送给视频解码装置(例如解码器)，视频解码装置再通过DLT将索引残差值转化为深度残差值，最后将深度残差值加到相应区域(例如预测图像块)中每个像素点的像素值上得到重建图像块。
DC(Direct Component，直流分量)是对待处理深度图像进行SDC处理时用于DLT变换的参数，SDC中需要当前深度图像块的原始DC(Original DC)及预测DC(Prediction DC)结合DLT得到对应索引值，进而得到索引残差值。深度图像的不同像素值很少。通过建立DLT，可以有效减少编码信号比特数。在对深度图像进行SDC的传统方法中，对于当前预测图像块，利用预测图像块的左上角、右上角、左下角、右下角这四个角像素点的四个像素值的均值作为预测DC。因为在SDC编解码过程中，预测DC起到协助计算深度残差值的作用，深度残差值代表预测图像块与原始图像块的整体偏差，深度残差值预测的精确与否直接关系到生成重建图像块时的编码性能优劣。然而，DLT变换会产生变换误差，影响深度残差值预测的精确度，进而影响到生成重建图像块时的编码性能优劣。所以有了一些试图减少变换误差的方法，以求提高生成重建图像块时的编码性能，然而均未能获得编码增益，编码增益体现生成重建图像块时编码性能的提升程度。
所以，对深度图像进行SDC的传统方法的现有改进技术，均不能减少DLT产生的变换误差，无法提升在生成深度图像的重建图像块时的编码性能。
发明内容
本发明的实施例提供一种图像处理方法、装置及系统，能减少对深度图进行SDC图像处理时DLT产生的变换误差，从而提升了图像处理结果的编码性能。
为达到上述目的，本发明的实施例采用如下技术方案：
第一方面，提供一种图像处理方法，包括：
获取待编码深度图像的原始图像块、至少一种图像处理模式和深 度查询表，其中，所述深度查询表包含至少一个索引值以及每个索引值对应的像素值，所述索引值用于表示所述索引值自身对应的像素值；
根据所述至少一种图像处理模式，获取每种图像处理模式各自对应的预测图像块；
根据所述预测图像块中一个像素点的像素值获取预测直流分量；
根据所述深度查询表和所述预测直流分量获得预测索引值；
根据所述原始图像块获取原始直流分量；
根据所述深度查询表和所述原始直流分量获得原始索引值；
将所述原始索引值减所述预测索引值得到索引残差值；
根据所述预测图像块和预设规则从所述至少一种图像处理模式中选取优选的图像处理模式；
对所述优选的图像处理模式、所述优选的图像处理模式对应的索引残差值和所述深度查询表进行编码以得到编码数据。
结合第一方面，在第一种可能的实现方式中，根据所述预测图像块中一个像素点的像素值获取预测直流分量，具体为：
根据所述预测图像块中右下角像素点的像素值获取预测直流分量；
或者，
根据所述预测图像块中心区域的一个像素点的像素值获取预测直流分量，所述预测图像块中心区域是所述预测图像块中心处边长为2个像素的正方形图像区域。
结合第一方面或第一种可能的实现方式，在第二种可能的实现方式中，根据所述预测图像块中一个像素点的像素值获取预测直流分量，包括：
在所述图像处理模式为角度模式时，根据所述预测图像块中一个像素点的像素值获取预测直流分量；
结合第一方面或第一种或第二种可能的实现方式中的任一种，在第三种可能的实现方式中，根据所述预测图像块中一个像素点的像素值获取预测直流分量，包括：
在所述图像处理模式为平面模式时，根据所述预测图像块中一个像素点的像素值获取预测直流分量。
结合第一方面或第一种至第三种可能的实现方式中的任一种，在第四种可能的实现方式中，根据所述预测图像块中一个像素点的像素值获取预测直流分量，包括：
在所述图像处理模式为直流分量模式时，根据所述预测图像块中一个像素点的像素值获取预测直流分量。
结合第一方面或第一种至第四种可能的实现方式中的任一种，在第五种可能的实现方式中，根据所述原始图像块获取原始直流分量，具体包括：
根据所述原始图像块中一个像素点的像素值获取原始直流分量。
结合第一方面和第五种可能的实现方式，在第六种可能的实现方式中，根据所述原始图像块中一个像素点的像素值获取原始直流分量，具体包括：
根据所述原始图像块中右下角像素点的像素值获取原始直流分量；
或者，
根据所述原始图像块中心区域的一个像素点的像素值获取原始直流分量，所述原始图像块中心区域是所述原始图像块中心处边长为2个像素的正方形图像区域。
结合第一方面或第一种至第六种可能的实现方式中的任一种，在第七种可能的实现方式中，根据所述至少一种图像处理模式，获取每种图像处理模式各自对应的预测图像块，具体包括：
在关闭帧内参考像素滤波器时，根据所述至少一种图像处理模式，获取每种图像处理模式各自对应的预测图像块。
结合第一方面或第一种至第七种可能的实现方式中的任一种，在第八种可能的实现方式中，根据所述预测图像块和预设规则从所述至少一种图像处理模式中选取优选的图像处理模式，具体包括：
根据所述预测图像块计算所述至少一种图像处理模式中与所述预测图像块对应的图像处理模式的率失真代价；
根据所述率失真代价，将所述至少一种图像处理模式中率失真代价最小的图像处理模式作为优选的图像处理模式。
第二方面，提供一种图像处理方法，包括：
获取待重建深度图像的编码数据，所述编码数据包括图像处理模式、索引残差值和深度查询表，其中，所述深度查询表包含至少一个索引值以及每个索引值对应的像素值，所述索引值用于表示所述索引值自身对应的像素值；
根据所述图像处理模式，获取预测图像块；
根据所述预测图像块中一个像素点的像素值获取预测直流分量；
根据所述深度查询表和所述预测直流分量获得预测索引值；
将所述索引残差值加所述预测索引值得到重建索引值；
根据所述深度查询表和所述重建索引值获得重建像素值，将所述重建像素值作为重建直流分量；
将所述重建直流分量减所述预测直流分量得到深度残差值；
为所述预测图像块的每一个像素点的像素值加上所述深度残差值得到重建图像块，所述重建图像块用于重建所述待重建深度图像。
结合第二方面，在第一种可能的实现方式中，根据所述预测图像块中一个像素点的像素值获取预测直流分量，具体为：
根据所述预测图像块中右下角像素点的像素值获取预测直流分量；
或者，
根据所述预测图像块中心区域的一个像素点的像素值获取预测直流分量，所述预测图像块中心区域是所述预测图像块中心处边长为2个像素的正方形图像区域。
结合第二方面或第一种可能的实现方式，在第二种可能的实现方式中，根据所述预测图像块中一个像素点的像素值获取预测直流分量，包括：
在所述图像处理模式为角度模式时，根据所述预测图像块中一个像素点的像素值获取预测直流分量；
结合第二方面或第一种或第二种可能的实现方式，在第三种可能的实现方式中，根据所述预测图像块中一个像素点的像素值获取预测直流分量，包括：
在所述图像处理模式为平面模式时，根据所述预测图像块中一个像素点的像素值获取预测直流分量。
结合第二方面或第一种至第三种可能的实现方式中的任一种，在第四种可能的实现方式中，根据所述预测图像块中一个像素点的像素值获取预测直流分量，包括：
在所述图像处理模式为直流分量模式时，根据所述预测图像块中一个像素点的像素值获取预测直流分量。
结合第二方面或第一种至第四种可能的实现方式中的任一种，在第种可能的实现方式中，根据所述图像处理模式，获取预测图像块，具体包括：
在关闭帧内参考像素滤波器时，根据所述图像处理模式，获取预测图像块。
第三方面，提供一种视频编码装置，包括：
获取单元，用于获取待编码深度图像的原始图像块、至少一种图像处理模式和深度查询表，其中，所述深度查询表包含至少一个索引值以及每个索引值对应的像素值，所述索引值用于表示所述索引值自身对应的像素值；
所述获取单元还用于根据所述获取单元获取的所述至少一种图像处理模式，获取每种图像处理模式各自对应的预测图像块；
预测直流分量单元，用于根据所述获取单元获取的所述预测图像块中一个像素点的像素值获取预测直流分量；
预测索引值单元，用于根据所述获取单元获取的所述深度查询表和所述直流分量单元获取的所述预测直流分量获得预测索引值。
原始直流分量单元，用于根据所述获取单元获取的所述原始图像块获取原始直流分量；
原始索引值单元，用于根据所述获取单元获取的所述深度查询表和所述原始直流分量单元获取的所述原始直流分量获得原始索引值；
索引残差值单元，用于将所述原始索引值单元获取的所述原始索引值减所述预测索引值单元获取的所述预测索引值得到索引残差值；
优选单元，用于根据所述获取单元获取的所述预测图像块和预设规则从所述至少一种图像处理模式中选取优选的图像处理模式；
编码单元，用于对所述优选单元选取的所述优选的图像处理模式、所述索引残差值单元获取的所述优选的图像处理模式对应的索引残差值和所述获取单元获取的所述深度查询表进行编码以得到编码数据。
结合第三方面，在第一种可能的实现方式中，所述预测直流分量单元具体用于根据所述获取单元获取的所述预测图像块中右下角像素点的像素值获取预测直流分量；
或者，
所述预测直流分量单元具体用于根据所述获取单元获取的所述预测图像块中心区域的一个像素点的像素值获取预测直流分量，所述预测图像块中心区域是所述预测图像块中心处边长为2个像素的正方形图像区域。
结合第三方面或第一种可能的实现方式，在第二种可能的实现方式中，所述预测直流分量单元，还用于在所述获取单元获取的所述图像处理模式为角度模式时，根据所述获取单元获取的所述预测图像块中一个像素点的像素值获取预测直流分量；
结合第三方面或第一种或第二种可能的实现方式中的任一种，在第三种可能的实现方式中，所述预测直流分量单元，还用于在所述获取单元获取的所述图像处理模式为平面模式时，根据所述获取单元获取的所述预测图像块中一个像素点的像素值获取预测直流分量。
结合第三方面或第一种至第三种可能的实现方式中的任一种，在第四种可能的实现方式中，所述预测直流分量单元，还用于在所述获取单元获取的所述图像处理模式为直流分量模式时，根据所述获取单元获取的所述预测图像块中一个像素点的像素值获取预测直流分量。
结合第三方面或第一种至第四种可能的实现方式中的任一种，在第五种可能的实现方式中，所述原始直流分量单元具体用于根据所述获 取单元获取的所述原始图像块中一个像素点的像素值获取原始直流分量。
结合第三方面和第五种可能的实现方式，在第六种可能的实现方式中，所述原始直流分量单元具体用于根据所述获取单元获取的所述原始图像块中右下角像素点的像素值获取原始直流分量；
或者，
所述原始直流分量单元具体用于根据所述获取单元获取的所述原始图像块中心区域的一个像素点的像素值获取原始直流分量，所述原始图像块中心区域是所述原始图像块中心处边长为2个像素的正方形图像区域。
结合第三方面或第一种至第六种可能的实现方式中的任一种，在第七种可能的实现方式中，所述获取单元具体用于在关闭帧内参考像素滤波器时，根据所述获取单元获取的所述至少一种图像处理模式，获取每种图像处理模式各自对应的预测图像块。
结合第三方面或第一种至第七种可能的实现方式中的任一种，在第八种可能的实现方式中，所述优选单元，具体包括：
率失真子单元，用于根据所述获取单元获取的所述预测图像块计算所述至少一种图像处理模式中与所述预测图像块对应的图像处理模式的率失真代价；
优选子单元，用于根据所述率失真子单元计算的所述率失真代价，将所述获取单元获取的所述至少一种图像处理模式中率失真代价最小的图像处理模式作为优选的图像处理模式。
第四方面，提供一种视频解码装置，包括：
获取单元，用于获取待重建深度图像的编码数据，所述编码数据包括图像处理模式、索引残差值和深度查询表，其中，所述深度查询表包含至少一个索引值以及每个索引值对应的像素值，所述索引值用于表示所述索引值自身对应的像素值；
所述获取单元还用于根据所述获取单元获取的所述图像处理模式，获取预测图像块；
预测直流分量单元，用于根据所述获取单元获取的所述预测图像块中一个像素点的像素值获取预测直流分量；
预测索引值单元，用于根据所述获取单元获取的所述深度查询表和所述直流分量单元获取的所述预测直流分量获得预测索引值。
重建索引值单元，用于将所述获取单元获取的所述索引残差值加所述预测索引值单元获取的所述预测索引值得到重建索引值；
重建直流分量单元，用于根据所述获取单元获取的所述深度查询表和所述重建索引值单元得到的所述重建索引值获得重建像素值，将所 述重建像素值作为重建直流分量；
深度残差值单元，用于将所述重建直流分量单元获得的所述重建直流分量减所述预测直流分量得到深度残差值；
重建单元，用于为所述获取单元获取的所述预测图像块的每一个像素点的像素值加上所述深度残差值单元得到的所述深度残差值得到重建图像块，所述重建图像块用于重建所述待重建深度图像。
结合第四方面，在第一种可能的实现方式中，所述预测直流分量单元具体用于根据所述获取单元获取的所述预测图像块中右下角像素点的像素值获取预测直流分量；
或者，
所述预测直流分量单元具体用于根据所述获取单元获取的所述预测图像块中心区域的一个像素点的像素值获取预测直流分量，所述预测图像块中心区域是所述预测图像块中心处边长为2个像素的正方形图像区域。
结合第四方面或第一种可能的实现方式，在第二种可能的实现方式中，所述预测直流分量单元，还用于在所述获取单元获取的所述图像处理模式为角度模式时，根据所述获取单元获取的所述预测图像块中一个像素点的像素值获取预测直流分量。
结合第四方面或第一种或第二种可能的实现方式，在第三种可能的实现方式中，所述预测直流分量单元，还用于在所述获取单元获取的所述图像处理模式为平面模式时，根据所述获取单元获取的所述预测图像块中一个像素点的像素值获取预测直流分量。
结合第四方面或第一种至第三种可能的实现方式中的任一种，在第四种可能的实现方式中，所述预测直流分量单元，还用于在所述获取单元获取的所述图像处理模式为直流分量模式时，根据所述获取单元获取的所述预测图像块中一个像素点的像素值获取预测直流分量。
结合第四方面或第一种至第四种可能的实现方式中的任一种，在第五种可能的实现方式中，所述获取单元具体用于在关闭帧内参考像素滤波器时，根据所述获取单元获取的所述图像处理模式，获取预测图像块。
本发明实施例提供的图像处理方法、装置及系统，能通过视频编码装置，获取待编码深度图像的原始图像块、至少一种图像处理模式和深度查询表，根据图像处理模式获取预测图像块，并根据这些获取到的信息得出原始直流分量，结合具体的图像处理模式根据所述预测图像块中一个像素点的像素值获取预测直流分量，根据原始直流分量和预测直流分量进行DLT计算出索引残差值，以便根据索引残差值重建被编码的深度图像。并且，能通过视频解码装置，获取待重建深度图像的编码数据，所述编码数据包括图像处理模式、索引残差值和深度查询表，并结 合具体的图像处理模式根据所述预测图像块中一个像素点的像素值获取预测直流分量，之后进行DLT变换，并结合索引残差值和DLT逆变换获取重建图像块，以便最终获得重建图像。因只使用一个像素点的像素值作为直流分量，相对现有技术中使用多个像素点的像素值的平均值作为直流分量而言，提高了直流分量的值完全由有效像素值得出的概率(多个像素点的像素值的平均值作为直流分量时，若平均值为小数则会对数值进行近似处理，会产生更多误差。同时，只有在所有参考像素点的像素值为有效像素值时，使用根据这些像素值得出的直流分量才不会产生DLT引起的变化误差，单个像素值是有效像素值的概率远大于多个像素值均是有效像素的概率)，所以减少了SDC的编码过程中DLT产生的变换误差，从而提升了图像处理结果的编码性能。
附图说明
为了更清楚地说明本发明实施例或现有技术中的技术方案，下面将对实施例或现有技术描述中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图仅仅是本发明的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据这些附图获得其他的附图。
图1为本发明的实施例提供的一种图像处理方法的流程示意图；
图2为本发明的实施例提供的一种图像处理模式的示意图；
图3为本发明的实施例提供的一种深度查询表的应用示意图；
图4为本发明的实施例提供的另一种深度查询表的应用示意图；
图5为本发明的又一实施例提供的一种图像处理方法的流程示意图；
图6为本发明的另一实施例提供的一种图像处理方法的流程示意图；
图7为本发明的另一个实施例提供的一种获取预测直流分量的示意图；
图8为本发明的再一个实施例提供的一种图像处理方法的流程示意图；
图9为本发明的实施例提供的一种视频编码装置的结构示意图；
图10为本发明的实施例提供的又一种视频编码装置的结构示意图；
图11为本发明的实施例提供的一种视频解码装置的结构示意图；
图12为本发明的又一个实施例提供的一种视频编码装置的结构示意图；
图13为本发明的又一个实施例提供的一种视频解码装置的结构示意图。
具体实施方式
下面将结合本发明实施例中的附图，对本发明实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅仅是本发明一部分实施例，而不是全部的实施例。基于本发明中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其他实施例，都属于本发明保护的范围。
本发明实施例提供一种图像处理系统，应用于多媒体领域，具体可以用于对深度图像的编码处理和解码处理。可以应用于在3D-HEVC中对深度图的SDC处理。所述图像处理系统包括能够相互连接进行数据通信的视频编码装置和视频解码装置。
在对深度图像的SDC处理中，视频编码装置需要先选取深度图像的待编码图像块(如原始图像块)和参照图像块(如预测图像块)，然后分别获取原始图像块的原始直流分量和预测图像块的预测直流分量，并结合已经建立好的DLT(Depth Lookup Table，深度查询表)对预测直流分量进行变换得到预测索引值，并对和原始直流分量进行变换得到原始索引值，计算出原始索引值和预测索引值之间的索引残差值，并将所述索引残差值编码到编码数据中。
之后，视频解码装置解码所述编码数据得到所述索引残差值，并使用与所述视频编码端装置相同的预测图像块选取模式得到预测图像块，得出预测直流分量，结合DLT得出预测索引值，将所述索引残差值与预测索引值相加得到重建索引值，再结合DLT进行逆变换得到重建直流分量，并用重建直流分量与预测直流分量相减得到深度残差值，将所述深度残差值与预测图像块的每一个像素点的像素值相加得到重建图像块。以上是一个对深度图像进行SDC处理的过程。
其中，SDC是应用在3D-HEVC中的一种残差数据编码方法。SDC可以应用在深度图像的所有帧内预测模式中，帧内预测模式是利用每一帧待处理图像自身对所述待处理图像进行预测编码的图像处理模式，其中包括HEVC帧内预测模式。HEVC帧内预测模式又包括直流分量模式、planar(平面)模式与33种角度模式。使用SDC对深度图像进行图像处理时，将深度图像中每个PU(Prediction Unit，预测单元)作为一个区域，该区域的尺寸总为2N×2N，单位为像素，N为大于1的整数。每个区域(例如预测图像块)在视频编码装置(例如编码器)通过建立DLT(Depth Lookup Table，深度查询表)编码一个索引残差值到编码数据中发送给视频解码装置(例如解码器)，视频解码装置再通过DLT将索引残差值转化为深度残差值，最后将深度残差值加到相应区域(例如预测图像块)中每个像素点的像素值上得到重建图像块。
DC(Direct Component，直流分量)是对待处理深度图像进行SDC处理时用于DLT变换的参数，SDC中需要当前深度图像块的原始DC (Original DC)及预测DC(Prediction DC)输入到DLT中得到对应索引值，进而得到索引残差值。深度图像的不同像素值很少。通过建立DLT，可以有效减少编码信号比特数。例如：深度像素值的范围为0-7，其中0、4、7为在原始深度图像中存在的像素值；1、2、3、5、6为在原始深度图像中不存在的像素值，直接编码深度值0、4、7需要7个编码，二进制中需要3位(23＝8)编码。通过DLT映射后，只需要编码0、4、7对应的3个编码，此时二进制中只需要2位(22＝4)编码，因此DLT可以减少编码比特数。然而，结合DLT进行的变换是不可逆的。由非有效像素获得的直流分量经DLT变换得到索引值后，因在DLT中无直接对应的索引值，所以会产生误差，无法再经逆变换重建出非有效像素值，所以直流分量与原始像素值相同的概率大小，直接影响到对单个图像块的DLT变换过程中误差产生概率的大小，也就影响对整幅图像的DLT变换过程中总误差的大小。
其中，所述视频编码装置可以用于对待处理图像(比如深度图)进行编码处理得到编码数据，所述视频解码装置可以用于所述对待处理图像的编码数据进行解码处理，完成图像重建。
所述视频编码装置可以与所述视频解码装置分处于不同的装置中，也可以处于同一装置中。或者所述视频编码装置可以与所述视频解码装置为同一装置，该装置既能够实现视频编码装置的编码功能，也可以实现所述视频解码装置的解码功能。
参照图1所示，本发明实施例提供一种图像处理方法，应用于多媒体领域，具体可以用于对深度图像的编码处理，可以应用于本发明上述实施例提供的图像处理系统中的视频编码装置，包括以下步骤：
101、获取待编码深度图像的原始图像块、至少一种图像处理模式和深度查询表。
其中，所述深度查询表包含至少一个索引值以及每个索引值对应的像素值，所述索引值用于表示所述索引值自身对应的像素值，所述深度查询表可以是预设的；所述至少一种图像处理模式可以包括平面模式、直流分量模式和角度模式中的一种或多种，所述角度模式包括33种角度模式。所述至少一种图像处理模式也可以包括这35种图像处理模式之外的图像处理模式。例如，所述至少一种图像处理模式，可以包括平面模式、直流分量模式和33种角度模式这35种图像处理模式，也可以只包括33种角度模式。在编码过程中，使用所述至少一种图像处理模式中的每种模式对待编码深度图像进行编码，结合后续步骤选取符合预设规则的图像处理模式。示例性的，参照图2所示，所述角度模式包括33种角度模式。所述原始图像块是边长为2N像素的正方形图像块，N为正偶数。
102、根据所述至少一种图像处理模式，获取每种图像处理模式各 自对应的预测图像块。
其中，所述预测图像块是边长为2N个像素的正方形图像块，N为正偶数；每种图像处理模式各自对应的预测图像块是预定的。
可选的，可以在关闭帧内参考像素滤波器时，根据所述至少一种图像处理模式，获取每种图像处理模式各自对应的预测图像块。
这样做的原因是，相对于所述帧内参考像素滤波器功能的状态为开启状态，帧内参考像素滤波器功能的状态为关闭状态时能获得更高的编码增益。
因为预测图像块是通过当前原始图像块的左一列和上一行参考像素预测得到的，所以参考像素的值直接影响着本发明的性能。由DLT建表过程可知，最初映射的深度像素值越接近原始像素值，DLT变换的误差就越小。帧内参考像素滤波器平滑了参考像素的值，会影响使用单个参考像素值作为预测DC的效果(不利于降低DLT变换引起的误差)，所以在关闭帧内参考像素滤波器时，能获得更高的增益。
103、根据所述预测图像块中一个像素点的像素值获取预测直流分量。
具体的，可以根据所述预测图像块中右下角像素点的像素值获取预测直流分量；或者，可以根据所述预测图像块中心区域的一个像素点的像素值获取预测直流分量，所述预测图像块中心区域是所述预测图像块中心处边长为2个像素的正方形图像区域。
因使用了单一像素的像素值作为预测直流分量，避免了对像素值平均带来的大概率DLT变换误差，可有效的减少DLT产生的变换误差，从而提升了在生成深度图像的重建图像块时的编码性能。
可选的，在所述图像处理模式为角度模式时，根据所述预测图像块中一个像素点的像素值获取预测直流分量；
进一步可选的，在所述图像处理模式为平面模式时，根据所述预测图像块中一个像素点的像素值获取预测直流分量。
更进一步可选的，在所述图像处理模式为直流分量模式时，根据所述预测图像块中一个像素点的像素值获取预测直流分量。
本步骤的更进一步的可选方案相对于本步骤的其他方案，因使更多的图像处理模式使用了单一像素的像素值作为预测直流分量，于是更有效的避免了对像素值平均带来的大概率DLT变换误差，有效的减少对深度图进行SDC图像处理时DLT产生的变换误差，从而提升了图像处理结果的编码性能，是本步骤的最佳可选方案。
另外，在所述图像处理模式为所述平面模式或所述直流分量模式时，还可以将所述预测图像块四个角像素点的像素值的均值作为预测直流分量。
这种另外的可选方案，是结合现有技术的改进方案，也能减少对深度图进行SDC图像处理时DLT产生的变换误差，从而提升了图像处理结果的编码性能，并且因和现有技术结合，利于本发明推广。
这里需要注意的是，本步骤中获取所述预测直流分量的以上的可选方案，均是根据实际需求提前预定的。
104、根据所述深度查询表和所述预测直流分量获得预测索引值。
具体的，可以参照图3所示：假设深度像素值的范围为D0-D7，其中D0、D4、D7为在深度图像中存在的像素值；D1、D2、D3、D5、D6为在原始深度图像中不存在的像素值，此时预设的深度查询表中仅有与像素值D0、D4、D7对应的索引值I0、I1、I2。若预测直流分量的值对应的像素值为D0、D4、D7，则获得像素值为D0、D4、D7时各自对应的索引值I0、I1、I2作为预测直流分量对应的索引值。同时，还可以参照图4所示：若预测直流分量不为D0、D4、D7，是D1、D2、D3、D5、D6，则按照预定规则选择临近像素值的预测索引值作为该预测直流分量对应的预测索引值。
105、根据所述原始图像块获取原始直流分量。
具体的，可以根据所述原始图像块中一个像素点的像素值获取原始直流分量。
更为具体的，可以根据所述原始图像块中右下角像素点的像素值获取原始直流分量；或者，根据所述原始图像块中心区域的一个像素点的像素值获取原始直流分量，所述原始图像块中心区域是所述原始图像块中心处边长为2个像素的正方形图像区域。
这样做的好处，类似于所述步骤103中各相似可选方案的有益效果。在此不再赘述。同时，结合各种可选方案，原始直流分量的获取方式中像素点的选取，可以参照获取所述预测直流分量选取像素点的位置，也可以不参照，如果获取所述原始直流分量参考的像素点位与获取所述预测直流分量参考的像素点位相同，则SDC会有最好的编码性能。
另外，还可以使用其他常用的方法获取所述原始直流分量，这样有利于本发明的实际推广。
这里需要注意的是，本步骤中获取所述原始直流分量的以上的可选方案，均是根据实际需求提前预定的。
106、根据所述深度查询表和所述原始直流分量获得原始索引值。
具体的，可以参照所述步骤104中获取预测索引值的方式来获取原始索引值，只需将预测直流分量替换为原始直流分量，预测索引值替换为原始索引值即可，其他部分相同，在此不再赘述。
107、将所述原始索引值减所述预测索引值得到索引残差值。
具体的，可以依照公式DeltaDC＝I(OrigDC)-I(PredDC)得到所述索引残差值，其中，I(OrigDC)为所述原始索引值，I(OrigDC)为所述预测索引值，DeltaDC为所述索引残差值，OrigDC为所述原始直流分量。
108、根据所述预测图像块和预设规则从所述至少一种图像处理模式中选取优选的图像处理模式。
具体的，可以根据所述预测图像块计算所述至少一种图像处理模式中与所述预测图像块对应的图像处理模式的率失真代价；然后，根据所述率失真代价，将所述至少一种图像处理模式中率失真代价最小的图像处理模式作为优选的图像处理模式。
比如，所述至少一种图像处理模式包括所述平面模式、所述直流分量模式和33种角度模式这一共35种图像处理模式，如果经计算，所述直流分量模式带来的率失真代价最小，则此时选择所述直流分量模式作为优选的图像处理模式。
109、对所述优选的图像处理模式、所述优选的图像处理模式对应的索引残差值和所述深度查询表进行编码以得到编码数据。
可选的，还可以发送编码数据，以便根据所述编码数据重建深度图像，所述编码数据包括所述优选的图像处理模式、所述优选的图像处理模式对应的索引残差值和所述深度查询表
本发明实施例提供的图像处理方法，能通过视频编码装置，获取待编码深度图像的原始图像块、至少一种图像处理模式和深度查询表，根据图像处理模式获取预测图像块，并根据这些获取到的信息得出原始直流分量，结合具体的图像处理模式根据所述预测图像块中一个像素点的像素值获取预测直流分量，根据原始直流分量和预测直流分量进行DLT计算出索引残差值，以便根据索引残差值重建被编码的深度图像。因只使用一个像素点的像素值作为直流分量，相对现有技术中使用多个像素点的像素值的平均值作为直流分量而言，提高了直流分量的值完全由有效像素值得出的概率(多个像素点的像素值的平均值作为直流分量时，若平均值为小数则会对数值进行近似处理，会产生更多误差。同时，只有在所有参考像素点的像素值为有效像素值时，使用根据这些像素值得出的直流分量才不会产生DLT引起的变化误差，单个像素值是有效像素值的概率远大于多个像素值均是有效像素的概率)，所以减少了SDC的编码过程中DLT产生的变换误差，从而提升了图像处理结果的编码性能。
参照图5所示，本发明实施例提供一种图像处理方法，应用于多媒体领域，具体可以用于对深度图像的图像处理，可以应用于本发明上述实施例提供的图像处理系统中的解码端，包括以下步骤：
501、获取待重建深度图像的编码数据，所述编码数据包括图像处 理模式、索引残差值和深度查询表。
其中，所述深度查询表包含至少一个索引值以及每个索引值对应的像素值，所述索引值用于表示所述索引值自身对应的像素值，所述深度查询表可以是预设的；所述图像处理模式可以包括平面模式、直流分量模式和角度模式中的一种，所述角度模式包括33种角度模式。所述图像处理模式也可以是这35种图像处理模式之外的一种图像处理模式。示例性的，参照图2所示，所述角度模式包括33种角度模式。所述原始图像块是边长为2N像素的正方形图像块，N为正偶数。
502、根据所述图像处理模式，获取预测图像块。
其中，所述预测图像块是边长为2N个像素的正方形图像块，N为正偶数；每种图像处理模式各自对应的预测图像块是预定的。
可选的，可以在关闭帧内参考像素滤波器时，根据所述图像处理模式，获取预测图像块。
这样做的原因是，相对于所述帧内参考像素滤波器功能的状态为开启状态，帧内参考像素滤波器功能的状态为关闭状态时能获得更高的编码增益。
因为预测图像块是通过待重建深度图像中待重建图像块的左一列和上一行参考像素预测得到的，所以参考像素的值直接影响着本发明的性能。由DLT建表过程可知，最初映射的深度像素值越接近原始像素值，DLT变换的误差就越小。帧内参考像素滤波器平滑了参考像素的值，会影响使用单个参考像素值作为预测DC的效果(不利于降低DLT变换引起的误差)，所以在关闭帧内参考像素滤波器时，能获得更高的增益。
503、根据所述预测图像块中一个像素点的像素值获取预测直流分量。
具体的，可以根据所述预测图像块中右下角像素点的像素值获取预测直流分量；或者，还可以根据所述预测图像块中心区域的一个像素点的像素值获取预测直流分量，所述预测图像块中心区域是所述预测图像块中心处边长为2个像素的正方形图像区域。
因使用了单一像素的像素值作为预测直流分量，避免了对像素值平均带来的大概率DLT变换误差，可有效的减少DLT产生的变换误差，从而提升了在生成深度图像的重建图像块时的编码性能。
可选的，在所述图像处理模式为角度模式时，根据所述预测图像块中一个像素点的像素值获取预测直流分量；
进一步可选的，在所述图像处理模式为平面模式时，根据所述预测图像块中一个像素点的像素值获取预测直流分量。
更进一步可选的，在所述图像处理模式为直流分量模式时，根据所述预测图像块中一个像素点的像素值获取预测直流分量。
本步骤的更进一步的可选方案相对于本步骤的其他方案，因使更多的图像处理模式使用了单一像素的像素值作为预测直流分量，于是更有效的避免了对像素值平均带来的大概率DLT变换误差，有效的减少对深度图进行SDC图像处理时DLT产生的变换误差，从而提升了图像处理结果的编码性能，是本步骤的最佳可选方案。
另外，在所述图像处理模式为所述平面模式或所述直流分量模式时，还可以将所述预测图像块四个角像素点的像素值的均值作为预测直流分量。
这种另外的可选方案，是结合现有技术的改进方案，也能减少对深度图进行SDC图像处理时DLT产生的变换误差，从而提升了图像处理结果的编码性能，并且因和现有技术结合，利于本发明推广。
这里需要注意的是，本步骤中获取所述预测直流分量的以上的可选方案，均是提前预定的，与图像处理系统中视频编码装置获取预测直流分量的预设方法相同。
504、根据所述深度查询表和所述预测直流分量获得预测索引值。
具体的，可以参照图3所示：假设深度像素值的范围为D0-D7，其中D0、D4、D7为在深度图像中存在的像素值；D1、D2、D3、D5、D6为在原始深度图像中不存在的像素值，此时预设的深度查询表中仅有与像素值D0、D4、D7对应的索引值I0、I1、I2。若预测直流分量的值对应的像素值为D0、D4、D7，则获得像素值为D0、D4、D7时各自对应的索引值I0、I1、I2作为预测直流分量对应的索引值。同时，还可以参照图4所示：若预测直流分量不为D0、D4、D7，是D1、D2、D3、D5、D6，则按照预定规则选择临近像素值的预测索引值作为该预测直流分量对应的预测索引值。
505、将所述索引残差值加所述预测索引值得到重建索引值。
506、根据所述深度查询表和所述重建索引值获得重建像素值，将所述重建像素值作为重建直流分量。
具体的，可以结合深度查询表，逆向参照所述步骤504中获取预测索引值的方式来逆向获取重建直流分量。即进行DLT逆变换。可以结合图4中所示的双向箭头连线，DLT逆变换时，只能根据重建索引值得到深度查询表中与索引值可以唯一对应的像素值作为重建直流分量。比如直流分量对应的像素值D1经DLT变换为索引值I0，而索引值I0只能逆变换为像素值D0对应的直流分量。
507、将所述重建直流分量减所述预测直流分量得到深度残差值。
具体的，可以参照公式DCOffset0＝I-1(I(PredDC)+DeltaDC)-PredDC获取所述深度残差值，其中DCOffset0为所述深度残差值，I(PredDC)为预测索引值，DeltaDC为索引残差值，I-1(I(PredDC)+DeltaDC)为所述重建直流分量(经DLT逆变换)，PredDC为所述预测直流分量。
508、为所述预测图像块的每一个像素点的像素值加上所述深度残差值得到重建图像块，所述重建图像块用于重建所述待重建深度图像。
本发明实施例提供的图像处理方法，能通过视频解码装置，获取待重建深度图像的编码数据，所述编码数据包括图像处理模式、索引残差值和深度查询表，并结合具体的图像处理模式根据所述预测图像块中一个像素点的像素值获取预测直流分量，之后进行DLT变换，并结合索引残差值和DLT逆变换获取重建图像块，以便最终获得重建图像。因只使用一个像素点的像素值作为直流分量，相对现有技术中使用多个像素点的像素值的平均值作为直流分量而言，提高了直流分量的值完全由有效像素值得出的概率(多个像素点的像素值的平均值作为直流分量时，若平均值为小数，而像素值不可为小数，则会产生更多误差。同时，只有在所有参考像素点的像素值为有效像素值时，使用根据这些像素值得出的直流分量才不会产生DLT引起的变化误差，单个像素值是有效像素值的概率远大于多个像素值均是有效像素的概率)，所以减少了SDC的编码过程中DLT产生的变换误差，从而提升了图像处理结果的编码性能。
在图1所示的实施例的基础上，参照图6所示，本发明实施例提供一种图像处理方法，应用于多媒体领域，具体可以用于对深度图像的图像处理，可以应用于本发明上述实施例提供的图像处理系统中的编码端，包括以下步骤：
601、获取待编码深度图像的原始图像块、至少一种图像处理模式和深度查询表。
其中，所述深度查询表包含至少一个索引值以及每个索引值对应的像素值，所述索引值用于表示所述索引值自身对应的像素值，所述深度查询表可以是预设的；所述至少一种图像处理模式可以包括平面模式、直流分量模式和角度模式中的一种或多种，所述角度模式包括33种角度模式。所述至少一种图像处理模式也可以包括这35种图像处理模式之外的图像处理模式。例如，所述至少一种图像处理模式，可以包括平面模式、直流分量模式和33种角度模式这35种图像处理模式，也可以只包括33种角度模式。在编码过程中，使用所述至少一种图像处理模式中的每种模式对待编码深度图像进行编码，结合后续步骤选取符合预设规则的图像处理模式。示例性的，参照图2所示，所述角度模式包括33种角度模式。所述原始图像块是边长为2N像素的正方形图像 块，N为正偶数。
602、在关闭帧内参考像素滤波器时，根据所述至少一种图像处理模式，获取每种图像处理模式各自对应的预测图像块。
其中，所述预测图像块是边长为2N个像素的正方形图像块，N为正偶数；每种图像处理模式各自对应的预测图像块是预定的。
这样做的原因是，相对于所述帧内参考像素滤波器功能的状态为开启状态，帧内参考像素滤波器功能的状态为关闭状态时能获得更高的编码增益。
因为预测图像块是通过当前原始图像块的左一列和上一行参考像素预测得到的，所以参考像素的值直接影响着本发明的性能。由DLT建表过程可知，最初映射的深度像素值越接近原始像素值，DLT变换的误差就越小。帧内参考像素滤波器平滑了参考像素的值，会影响使用单个参考像素值作为预测DC的效果(不利于降低DLT变换引起的误差)，所以在关闭帧内参考像素滤波器时，能获得更高的增益。
603、根据所述预测图像块中右下角像素点的像素值获取预测直流分量；
具体的，可以参照图7所示，根据第一预测等式PredDC＝PW-1,W-1获取预测直流分量，其中，PredDC为所述预测直流分量，PW-1,W-1为所述预测图像块中右下角像素点的像素值，W为所述预测图像块的边长数值。
或者，可选的，可以根据所述预测图像块中心区域的一个像素点的像素值获取预测直流分量，所述预测图像块中心区域是所述预测图像块中心处边长为2个像素的正方形图像区域。
具体的，可以参照图7所示，根据第二预测等式PredDC＝P(W-1)＞＞1,(W-1)＞＞1获取预测直流分量，其中，PredDC为所述预测直流分量，P(W-1)＞＞1,(W-1)＞＞1为所述预测图像块中心区域的左上角像素点的像素值，W为所述预测图像块的边长数值；
或者，根据第三预测等式PredDC＝P(W-1)＞＞1,W＞＞1获取预测直流分量，其中，PredDC为所述预测直流分量，P(W-1)＞＞1,W＞＞1为所述预测图像块中心区域的左下角像素点的像素值，W为所述预测图像块的边长数值；
或者，根据第四预测等式PredDC＝PW＞＞1,(W-1)＞＞1获取预测直流分量，其中，PredDC为所述预测直流分量，PW＞＞1,(W-1)＞＞1为所述预测图像块中心区域的右上角像素点的像素值，W为所述预测图像块的边长数值；
或者，根据第五预测等式PredDC＝PW＞＞1,W＞＞1获取预测直流分量，其 中，PredDC为所述预测直流分量，PW＞＞1,W＞＞1为所述预测图像块中心区域的右下角像素点的像素值，W为所述预测图像块的边长数值。
因使用了单一像素的像素值作为预测直流分量，避免了对像素值平均带来的大概率DLT变换误差，可有效的减少DLT产生的变换误差，从而提升了在生成深度图像的重建图像块时的编码性能。
另外，在所述图像处理模式为所述平面模式或所述直流分量模式时，还可以将所述预测图像块四个角像素点的像素值的均值作为预测直流分量。具体的，在所述图像处理模式为所述平面模式或所述直流分量模式时，可以参照如下公式Pred DC＝(P0,0+P(W-1),0+P0,(W-1)+P(W-1),(W-1)+2)＞＞2，获得预测直流分量，其中，Pred DC是预测直流分量，P0,0为左上角像素点的像素值，P(W-1),0为右上角像素点的像素值，P0,(W-1)为左下角像素点的像素值，P(W-1),(W-1)为右下角像素点的像素值，＞＞2表示计算机计算中的位移操作，是向右位移2位，因为计算机计算普遍使用的时二进制计算方式，所以向右位移2位相当于将Pred DC除以2的2次方，即除以4，此时因选取了4个像素点，这样的计算就相当于对4个像素点的像素值求均值。
这种另外的可选方案，是结合现有技术的改进方案，也能减少对深度图进行SDC图像处理时DLT产生的变换误差，从而提升了图像处理结果的编码性能，并且因和现有技术结合，利于本发明推广。
因为右下角像素值在33种角度模式中可以有33个不同位置的参考像素，中心点像素可以有16-17个不同位置的参考像素，这些位置的预测值的参考像素值的位置范围较广，在VSO进行RDO选择时就能选择到优选的模式(例如根据率失真代价进行对各图像处理模式的选择)。同时，单个像素值存在于原始深度图像中的可能性比均值存在于原始深度图像中的可能性要大很多，因此在DLT变换与反变换后引起的误差也就会小的多。
这里需要注意的是，本步骤中获取所述预测直流分量的以上的可选方案，均是根据实际需求提前预定的。
604、根据所述深度查询表和所述预测直流分量获得预测索引值。
具体的，可以参照图3所示：假设深度像素值的范围为D0-D7，其中D0、D4、D7为在深度图像中存在的像素值；D1、D2、D3、D5、D6为在原始深度图像中不存在的像素值，此时预设的深度查询表中仅有与像素值D0、D4、D7对应的索引值I0、I1、I2。若预测直流分量的值对应的像素值为D0、D4、D7，则获得像素值为D0、D4、D7时各自对应 的索引值I0、I1、I2作为预测直流分量对应的索引值。同时，还可以参照图4所示：若预测直流分量不为D0、D4、D7，是D1、D2、D3、D5、D6，则按照预定规则选择临近像素值的预测索引值作为该预测直流分量对应的预测索引值。
605、根据所述原始图像块中右下角像素点的像素值获取原始直流分量。
具体的，根据第一原始等式Orig DC＝PW-1,W-1获取原始直流分量，其中，OrigDC为所述原始直流分量，PW-1,W-1为所述原始图像块中右下角像素点的像素值，W为所述原始图像块的边长数值。
或者，可选的，结合所述步骤603具体选择的方案，本步骤还可以根据所述原始图像块中心区域的一个像素点的像素值获取原始直流分量，所述原始图像块中心区域是所述原始图像块中心处边长为2个像素的正方形图像区域。
具体的，根据第二原始等式OrigDC＝P(W-1)＞＞1,(W-1)＞＞1获取原始直流分量，其中，OrigDC为所述原始直流分量，P(W-1)＞＞1,(W-1)＞＞1为所述原始图像块中心区域的左上角像素点的像素值，W为所述原始图像块的边长数值；
或者，根据第三原始等式OrigDC＝P(W-1)＞＞1,W＞＞1获取原始直流分量，其中，OrigDC为所述原始直流分量，P(W-1)＞＞1,W＞＞1为所述原始图像块中心区域的左下角像素点的像素值，W为所述原始图像块的边长数值；
或者，根据第四原始等式OrigDC＝PW＞＞1,(W-1)＞＞1获取原始直流分量，其中，OrigDC为所述原始直流分量，PW＞＞1,(W-1)＞＞1为所述原始图像块中心区域的右上角像素点的像素值，W为所述原始图像块的边长数值；
或者，根据第五原始等式OrigDC＝PW＞＞1,W＞＞1获取原始直流分量，其中，OrigDC为所述原始直流分量，PW＞＞1,W＞＞1为所述原始图像块中心区域的右下角像素点的像素值，W为所述原始图像块的边长数值。
这样做的好处，类似于所述步骤103中各相似可选方案的有益效果。在此不再赘述。同时，结合各种可选方案，原始直流分量的获取方式中像素点的选取，可以参照获取所述预测直流分量选取像素点的位置，也可以不参照，如果获取所述原始直流分量参考的像素点位与获取所述预测直流分量参考的像素点位相同，则SDC会有最好的编码性能。
另外，还可以使用其他常用的方法获取所述原始直流分量，这样 有利于本发明的实际推广。
比如，如果所述原始图像块尺寸为64×64或者32×32(单位是像素)，则先对原始块进行2:1的下采样，再将下采样块的所有像素值的均值作为原始直流分量，即
Figure PCTCN2015077721-appb-000001
如果所述原始图像块尺寸小于32×32，则直接将原始块的所有像素值的均值作为原始直流分量，即
Figure PCTCN2015077721-appb-000002
此种可选方案可作为一种默认的可选方案，图像处理效果并非优选，但是最常用，对现有技术的改动最小，最利于推广。
这里需要注意的是，本步骤中获取所述原始直流分量的以上的可选方案，均是根据实际需求提前预定的。
606、根据所述深度查询表和所述原始直流分量获得原始索引值。
具体的，可以参照步骤604中获得预测索引值的方式获取原始索引值，只需将预测直流分量替换为原始直流分量，预测索引值替换为原始索引值即可，其他部分相同，在此不再赘述。
607、将所述原始索引值减所述预测索引值得到索引残差值。
具体的，可以依照公式DeltaDC＝I(OrigDC)-I(PredDC)得到所述索引残差值，其中，I(OrigDC)为所述原始索引值，I(OrigDC)为所述预测索引值，DeltaDC为所述索引残差值，OrigDC为所述原始直流分量。
608、根据所述预测图像块和预设规则从所述至少一种图像处理模式中选取优选的图像处理模式。
具体的，可以根据所述预测图像块计算所述至少一种图像处理模式中与所述预测图像块对应的图像处理模式的率失真代价；然后，根据所述率失真代价，将所述至少一种图像处理模式中率失真代价最小的图像处理模式作为优选的图像处理模式。
比如，所述至少一种图像处理模式包括所述平面模式、所述直流分量模式和33种角度模式这一共35种图像处理模式，如果经计算，所述直流分量模式带来的率失真代价最小，则此时选择所述直流分量模式作为优选的图像处理模式。
609、对所述优选的图像处理模式、所述优选的图像处理模式对应的索引残差值和所述深度查询表进行编码以得到编码数据。
可选的，还可以发送编码数据，以便根据所述编码数据重建深度图像，所述编码数据包括所述优选的图像处理模式、所述优选的图像处理模式对应的索引残差值和所述深度查询表。
本发明实施例提供的图像处理方法，能通过视频编码装置，获取待编码深度图像的原始图像块、至少一种图像处理模式和深度查询表，根据图像处理模式获取预测图像块，结合具体的图像处理模式根据所述预测图像块右下角像素点的像素值获取预测直流分量，并根据原始图像块右下角像素点的像素值得出原始直流分量，根据原始直流分量和预测直流分量进行DLT计算出索引残差值，以便根据索引残差值重建被编码的深度图像。因右下角像素点在33种角度模式中可更大的提高直流分量的值完全由有效像素值得出的概率，所以减少了SDC的编码过程中DLT产生的变换误差，能在VSO进行RDO选择时选择到更合适的编码结果，从而提升了图像处理结果的编码性能。
在图5所示的实施例的基础上，参照图8所示，本发明实施例提供一种图像处理方法，应用于多媒体领域，具体可以用于对深度图像的图像处理，可以应用于本发明上述实施例提供的图像处理系统中的解码端，包括以下步骤：
801、获取待重建深度图像的编码数据，所述编码数据包括图像处理模式、索引残差值和深度查询表。
其中，所述深度查询表包含至少一个索引值以及每个索引值对应的像素值，所述索引值用于表示所述索引值自身对应的像素值，所述深度查询表可以是预设的；所述图像处理模式可以包括平面模式、直流分量模式和角度模式中的一种，所述角度模式包括33种角度模式。所述图像处理模式也可以是这35种图像处理模式之外的一种图像处理模式。示例性的，参照图2所示，所述角度模式包括33种角度模式。所述原始图像块是边长为2N像素的正方形图像块，N为正偶数。
802、在关闭帧内参考像素滤波器时，根据所述图像处理模式，获取预测图像块。
其中，所述预测图像块是边长为2N个像素的正方形图像块，N为正偶数；每种图像处理模式各自对应的预测图像块是预定的。
这样做的原因是，相对于所述帧内参考像素滤波器功能的状态为开启状态，帧内参考像素滤波器功能的状态为关闭状态时能获得更高的编码增益。
因为预测图像块是通过待重建深度图像中待重建图像块的左一列和上一行参考像素预测得到的，所以参考像素的值直接影响着本发明的性能。由DLT建表过程可知，最初映射的深度像素值越接近原始像素值，DLT变换的误差就越小。帧内参考像素滤波器平滑了参考像素的值，会影响使用单个参考像素值作为预测DC的效果(不利于降低DLT变换引起的误差)，所以在关闭帧内参考像素滤波器时，能获得更高的增益。
803、根据所述预测图像块中右下角像素点的像素值获取预测直流分量；
具体的，可以参照图7所示，根据第一预测等式PredDC＝PW-1,W-1获取预测直流分量，其中，PredDC为所述预测直流分量，PW-1,W-1为所述预测图像块中右下角像素点的像素值，W为所述预测图像块的边长数值。
或者，可选的，还可以根据所述预测图像块中心区域的一个像素点的像素值获取预测直流分量，所述预测图像块中心区域是所述预测图像块中心处边长为2个像素的正方形图像区域。
具体的，可以参照图7所示，根据第二预测等式PredDC＝P(W-1)＞＞1,(W-1)＞＞1获取预测直流分量，其中，PredDC为所述预测直流分量，P(W-1)＞＞1,(W-1)＞＞1为所述预测图像块中心区域的左上角像素点的像素值，W为所述预测图像块的边长数值；
或者，根据第三预测等式PredDC＝P(W-1)＞＞1,W＞＞1获取预测直流分量，其中，PredDC为所述预测直流分量，P(W-1)＞＞1,W＞＞1为所述预测图像块中心区域的左下角像素点的像素值，W为所述预测图像块的边长数值；
或者，根据第四预测等式PredDC＝PW＞＞1,(W-1)＞＞1获取预测直流分量，其中，PredDC为所述预测直流分量，PW＞＞1,(W-1)＞＞1为所述预测图像块中心区域的右上角像素点的像素值，W为所述预测图像块的边长数值；
或者，根据第五预测等式PredDC＝PW＞＞1,W＞＞1获取预测直流分量，其中，PredDC为所述预测直流分量，PW＞＞1,W＞＞1为所述预测图像块中心区域的右下角像素点的像素值，W为所述预测图像块的边长数值。
因使用了单一像素的像素值作为预测直流分量，避免了对像素值平均带来的大概率DLT变换误差，可有效的减少DLT产生的变换误差，从而提升了在生成深度图像的重建图像块时的编码性能。
另外，在所述图像处理模式为所述平面模式或所述直流分量模式时，还可以将所述预测图像块四个角像素点的像素值的均值作为预测直流分量。具体的，在所述图像处理模式为所述平面模式或所述直流分量模式时，可以参照如下公式Pred DC＝(P0,0+P(W-1),0+P0,(W-1)+P(W-1),(W-1)+2)＞＞2，获得预测直流分量，其中，Pred DC是预测直流分量，P0,0为左上角像素点的像素值，P(W-1),0为右上角像素点的像素值，P0,(W-1)为左下角像素点的像素值，P(W-1),(W-1)为右下角像素点的像素值，＞＞2表示计算机计算中的位移操作，是向右位移2位，因为计算机计算普遍使用的时二进制计算方式，所以向右位移2位相当于将Pred DC除以2的2次方，即除以4，此时因选取了4 个像素点，这样的计算就相当于对4个像素点的像素值求均值。
这种另外的可选方案，是结合现有技术的改进方案，也能减少对深度图进行SDC图像处理时DLT产生的变换误差，从而提升了图像处理结果的编码性能，并且因和现有技术结合，利于本发明推广。
这些具体可选方案带来的好处是，右下角像素值在33种角度模式中可以有33个不同位置的参考像素，中心点像素可以有16-17个不同位置的参考像素，这些位置的预测值的参考像素值的位置范围较广，在VSO(View Synthesis Optimization，视点合成优化)进行RDO(Rate Distortion Optimization，率失真优化)选择时就能选择到优选的模式(例如根据率失真代价进行对各图像处理模式的选择)。同时，单个像素值存在于原始深度图像中的可能性比均值存在于原始深度图像中的可能性要大很多，因此在DLT变换与反变换后引起的误差也就会小的多。
这里需要注意的是，本步骤中获取所述预测直流分量的以上的可选方案，均是提前预定的，与图像处理系统中视频编码装置获取预测直流分量的预设方法相同。
804、根据所述深度查询表和所述预测直流分量获得预测索引值。
具体的，可以参照图3所示：假设深度像素值的范围为D0-D7，其中D0、D4、D7为在深度图像中存在的像素值；D1、D2、D3、D5、D6为在原始深度图像中不存在的像素值，此时预设的深度查询表中仅有与像素值D0、D4、D7对应的索引值I0、I1、I2。若预测直流分量的值对应的像素值为D0、D4、D7，则获得像素值为D0、D4、D7时各自对应的索引值I0、I1、I2作为预测直流分量对应的索引值。同时，还可以参照图4所示：若预测直流分量不为D0、D4、D7，是D1、D2、D3、D5、D6，则按照预定规则选择临近像素值的预测索引值作为该预测直流分量对应的预测索引值。
805、将所述索引残差值加所述预测索引值得到重建索引值。
806、根据所述深度查询表和所述重建索引值获得重建像素值，将所述重建像素值作为重建直流分量。
具体的，可以结合深度查询表，逆向参照所述步骤804中获取预测索引值的方式来逆向获取重建直流分量。即进行DLT逆变换。可以结合图4中所示的双向箭头连线，DLT逆变换时，只能根据重建索引值得到深度查询表中与索引值可以唯一对应的像素值作为重建直流分量。比如直流分量对应的像素值D1经DLT变换为索引值I0，而索引值I0只能逆 变换为像素值D0对应的直流分量。
807、将所述重建直流分量减所述预测直流分量得到深度残差值。
具体的，可以参照公式DCOffset0＝I-1(I(PredDC)+DeltaDC)-PredDC获取所述深度残差值，其中DCOffset0为所述深度残差值，I(PredDC)为预测索引值，DeltaDC为索引残差值，I-1(I(PredDC)+DeltaDC)为所述重建直流分量(经DLT逆变换)，PredDC为所述预测直流分量。
808、为所述预测图像块的每一个像素点的像素值加上所述深度残差值得到重建图像块，所述重建图像块用于重建所述待重建深度图像。
本发明实施例提供的图像处理方法，能通过视频解码装置，获取待重建深度图像的编码数据，所述编码数据包括图像处理模式、索引残差值和深度查询表，并结合具体的图像处理模式根据所述预测图像块中一个像素点的像素值获取预测直流分量，之后进行DLT变换，并结合索引残差值和DLT逆变换获取重建图像块，以便最终获得重建图像。因右下角像素点在33种角度模式中可更大的提高直流分量的值完全由有效像素值得出的概率，所以减少了SDC的编码过程中DLT产生的变换误差，能在VSO进行RDO选择时选择到更合适的编码结果，从而提升了图像处理结果的编码性能。
针对本发明以上各实施例提供的图像处理方法，为了更直观的说明本发明实施例获得的编码增益，以展现带来的编码性能提升，下面提供表1.1、表1.2、表2.1和表2.2、表3和表4用来说明所述步骤803中各种可替代方案带来的有益效果，表中左列是进行测试使用的测试视频的名称，这些视频是国际公认的标准测试视频：
表1.1、角度模式的预测图像块中心区域左下角像素值作为预测DC
Figure PCTCN2015077721-appb-000003
Figure PCTCN2015077721-appb-000004
表1.2、角度模式的预测图像块右下角像素值作为预测DC
Figure PCTCN2015077721-appb-000005
表1.1与1.2中展示了AI(All Intra,全部为帧内模式)2帧下将本发明只应用在深度图的33种角度模式中的测试结果，其中，表1.1表明的方案是采用预测块的中心块内左下角像素值作为预测DC，带来的平均编码增益是0.10％；表1.2表明的方案是采用预测块的右下角像素值作为预测DC，带来的平均编码增益是0.08％。
表2.1、DC模式与角度模式的预测图像块中心区域左下角像素值作为预测DC的结果
Figure PCTCN2015077721-appb-000006
Figure PCTCN2015077721-appb-000007
表2.2、planar模式与角度模式的预测图像块内右下角像素值作为预测DC的结果
Figure PCTCN2015077721-appb-000008
本发明对角度模式最有效，但是因为单个像素值预测会减少DLT变换的误差，所以在表1.1与表1.2应用情形的基础上，再扩大本发明的应用模式到DC模式与planar模式也会带来一定的增益。表2.1和表2.2展示了AI 2帧下的测试结果，其中，表2.1表示深度图的DC模式与角度模式采用中心区域内左下角像素作预测DC，planar模式采用4个角像素均值作预测DC的结果，带来的平均编码增益是0.15％；表2.2表示深度图的planar模式与角度模式采用预测块内右下角像素作预测的DC，DC模式采用4个角像素均值作预测DC的结果，带来的平均编码增益是0.11％。
结合本发明实施例的步骤803和之前实施例的步骤603、步骤605，下面提供表3和表4用来说明这三个步骤中各种可替代方案的组合带来的有益效果：
表3、所有HEVC帧内SDC模式的原始DC与预测DC都采用图像块右下角像素点的像素值
Figure PCTCN2015077721-appb-000009
Figure PCTCN2015077721-appb-000010
对于深度图的所有HEVC帧内模式SDC的原始DC与预测DC的计算都采用本发明技术，原始DC与预测DC都采用一个像素值作为DC，会比单独在预测DC中应用本发明带来更多的增益。表3展示了该方案AI2帧下的测试结果，对于深度图像的所有HEVC帧内SDC模式，原始DC与预测DC都采用右下角位置的像素值，此时的平均编码增益是0.29％。
表4、表3的设定条件下，关闭帧内参考像素滤波器功能
Figure PCTCN2015077721-appb-000011
Figure PCTCN2015077721-appb-000012
其中，表4展示了关闭帧内参考像素滤波器功能后再获取预测DC时表3设定条件下的实验结果，此时的平均编码增益是0.31％，表4所展示的技术方案为本发明的优选技术方案。
上述各表能展示二维视频编码时只需要编码一个视点的彩色视频；三维视频则需要分别编码多个视点的彩色及深度视频，不同视点的视频才能让观看者产生立体感受。不同视点的视频就相当于从不同方向看到的同一个场景的画面。深度视频反映场景的距离信息，不是用来直接观看的，而用来辅助彩色视频合成其他视点视频的。所以不会直接检验深度图像的编码质量，而是通过合成视点的质量来评价深度图像与彩色视频的编码质量。在国际标准测试视频中，同时编码传输三个视点的彩色与深度视频，再通过深度视频与彩色视频合成任意视点的视频，通过合成视频质量的好坏来衡量整体编码性能。表1.1与1.2中的video0、video1、video2分别表示三个不同视点的彩色视频，对应的实验数据表示对原始视频中的原始图像进行编码处理，再解码处理后得到的重建图像组成的重建视频与原始视频的PSNR(Peak Signal to Noise Ratio，峰值信噪比)。video只表示彩色视频，total表示彩色加上深度视频。video PSNR/video bitrate表示彩色视频的PSNR与彩色视频的比特数的比值；video PSNR/total bitrate表示彩色视频的PSNR与彩色视频与深度视频的比特数和的比值；synth PSNR/total bitrate表示合成视频的PSNR与彩色视频与深度视频的比特数和的比值，这个数值最能反映编码性能的好坏。
由上可知，本发明实施例提供的图像处理方法，在视频解码装置对编码数据的解码过程中，对任意所述图像处理模式，能根据所述预测图像块中一个像素点的像素值获取预测直流分量，且所述解码端使用的预测直流分量具体获取方式与所述编码端相同，这样就能减少解码过程中DLT产生的变换误差，从而提升了在生成深度图像的重建图像块时的编码性能。
参照图9所示，本发明实施例提供一种视频编码装置，应用于多媒体领域，具体可以用于对深度图像的图像处理，可以应用于本发明的实施例提供的图像处理系统，可以结合本发明的上述实施例为图像处理系统中视频编码装置提供的图像处理方法进行使用，包括以下结构：
获取单元901，用于获取待编码深度图像的原始图像块、至少一种 图像处理模式和深度查询表，其中，所述深度查询表包含至少一个索引值以及每个索引值对应的像素值，所述索引值用于表示所述索引值自身对应的像素值；
所述获取单元901还用于根据所述获取单元901获取的所述至少一种图像处理模式，获取每种图像处理模式各自对应的预测图像块；
预测直流分量单元902，用于根据所述获取单元901获取的所述预测图像块中一个像素点的像素值获取预测直流分量；
预测索引值单元903，用于根据所述获取单元901获取的所述深度查询表和所述直流分量单元获取的所述预测直流分量获得预测索引值。
原始直流分量单元904，用于根据所述获取单元901获取的所述原始图像块获取原始直流分量；
原始索引值单元905，用于根据所述获取单元901获取的所述深度查询表和所述原始直流分量单元904获取的所述原始直流分量获得原始索引值；
索引残差值单元906，用于将所述原始索引值单元905获取的所述原始索引值减所述预测索引值单元903获取的所述预测索引值得到索引残差值；
优选单元907，用于根据所述获取单元901获取的所述预测图像块和预设规则从所述至少一种图像处理模式中选取优选的图像处理模式；
编码单元908，用于对所述优选单元907选取的所述优选的图像处理模式、所述索引残差值单元906获取的所述优选的图像处理模式对应的索引残差值和所述获取单元901获取的所述深度查询表进行编码以得到编码数据。
所述视频编码装置还可以向外发送所述编码数据，以便其他装置根据所述编码数据生成重建深度图像。
可选的，所述预测直流分量单元902具体用于根据所述获取单元901获取的所述预测图像块中右下角像素点的像素值获取预测直流分量；
或者，所述预测直流分量单元902具体用于根据所述获取单元901获取的所述预测图像块中心区域的一个像素点的像素值获取预测直流分量，所述预测图像块中心区域是所述预测图像块中心处边长为2个像素的正方形图像区域。
可选的，所述预测直流分量单元902，还用于在所述获取单元901获取的所述图像处理模式为角度模式时，根据所述获取单元901获取的所述预测图像块中一个像素点的像素值获取预测直流分量。
可选的，所述预测直流分量单元902，还用于在所述获取单元901获取的所述图像处理模式为平面模式时，根据所述获取单元901获取的 所述预测图像块中一个像素点的像素值获取预测直流分量。
可选的，所述预测直流分量单元902，还用于在所述获取单元901获取的所述图像处理模式为直流分量模式时，根据所述获取单元901获取的所述预测图像块中一个像素点的像素值获取预测直流分量。
可选的，所述原始直流分量单元904具体用于根据所述获取单元901获取的所述原始图像块中一个像素点的像素值获取原始直流分量。
可选的，所述原始直流分量单元904具体用于根据所述获取单元901获取的所述原始图像块中右下角像素点的像素值获取原始直流分量；
或者，所述原始直流分量单元904具体用于根据所述获取单元901获取的所述原始图像块中心区域的一个像素点的像素值获取原始直流分量，所述原始图像块中心区域是所述原始图像块中心处边长为2个像素的正方形图像区域。
可选的，所述获取单元901具体用于在关闭帧内参考像素滤波器时，根据所述获取单元901获取的所述至少一种图像处理模式，获取每种图像处理模式各自对应的预测图像块。
可选的，参照图10所示，所述优选单元907，具体包括：
率失真子单元9071，用于根据所述获取单元901获取的所述预测图像块计算所述至少一种图像处理模式中与所述预测图像块对应的图像处理模式的率失真代价；
优选子单元9072，用于根据所述率失真子单元9071计算的所述率失真代价，将所述获取单元901获取的所述至少一种图像处理模式中率失真代价最小的图像处理模式作为优选的图像处理模式。
本发明实施例提供的视频编码装置，能获取待编码深度图像的原始图像块、至少一种图像处理模式和深度查询表，根据图像处理模式获取预测图像块，结合具体的图像处理模式根据所述预测图像块右下角像素点的像素值获取预测直流分量，并根据原始图像块右下角像素点的像素值得出原始直流分量，根据原始直流分量和预测直流分量进行DLT计算出索引残差值，以便根据索引残差值重建被编码的深度图像。因右下角像素点在33种角度模式中可更大的提高参考像素值和索引值的契合度，所以减少了SDC的编码过程中DLT产生的变换误差，能在VSO进行RDO选择时选择到更合适的编码结果，从而提升了图像处理结果的编码性能。
参照图11所示，本发明实施例提供一种视频解码装置，应用于多媒体领域，具体可以用于对深度图像的图像处理，可以应用于本发明的实施例提供的图像处理系统，可以结合本发明的上述实施例为图像处理系统中视频解码装置提供的图像处理方法进行使用，包括以下结构：
获取单元1101，用于获取待重建深度图像的编码数据，所述编码 数据包括图像处理模式、索引残差值和深度查询表，其中，所述深度查询表包含至少一个索引值以及每个索引值对应的像素值，所述索引值用于表示所述索引值自身对应的像素值；
所述获取单元1101还用于根据所述获取单元1101获取的所述图像处理模式，获取预测图像块；
预测直流分量单元1102，用于根据所述获取单元1101获取的所述预测图像块中一个像素点的像素值获取预测直流分量；
预测索引值单元1103，用于根据所述获取单元1101获取的所述深度查询表和所述直流分量单元获取的所述预测直流分量获得预测索引值。
重建索引值单元1104，用于将所述获取单元1101获取的所述索引残差值加所述预测索引值单元1103获取的所述预测索引值得到重建索引值；
重建直流分量单元1105，用于根据所述获取单元1101获取的所述深度查询表和所述重建索引值单元1104得到的所述重建索引值获得重建像素值，将所述重建像素值作为重建直流分量；
深度残差值单元1106，用于将所述重建直流分量单元1105获得的所述重建直流分量减所述预测直流分量得到深度残差值；
重建单元1107，用于为所述获取单元1101获取的所述预测图像块的每一个像素点的像素值加上所述深度残差值单元1106得到的所述深度残差值得到重建图像块，所述重建图像块用于重建所述待重建深度图像。
可选的，所述预测直流分量单元1102具体用于根据所述获取单元1101获取的所述预测图像块中右下角像素点的像素值获取预测直流分量；
或者，所述预测直流分量单元1102具体用于根据所述获取单元1101获取的所述预测图像块中心区域的一个像素点的像素值获取预测直流分量，所述预测图像块中心区域是所述预测图像块中心处边长为2个像素的正方形图像区域。
可选的，所述预测直流分量单元1102，还用于在所述获取单元1101获取的所述图像处理模式为角度模式时，根据所述获取单元1101获取的所述预测图像块中一个像素点的像素值获取预测直流分量。
可选的，所述预测直流分量单元1102，还用于在所述获取单元1101获取的所述图像处理模式为平面模式时，根据所述获取单元1101获取的所述预测图像块中一个像素点的像素值获取预测直流分量。
可选的，所述预测直流分量单元1102，还用于在所述获取单元1101获取的所述图像处理模式为直流分量模式时，根据所述获取单元1101获取的所述预测图像块中一个像素点的像素值获取预测直流分量。
可选的，所述获取单元1101具体用于在关闭帧内参考像素滤波器时，根据所述获取单元1101获取的所述图像处理模式，获取预测图像块。
本发明实施例提供的视频解码装置，能获取待重建深度图像的编码数据，所述编码数据包括图像处理模式、索引残差值和深度查询表，并结合具体的图像处理模式根据所述预测图像块中一个像素点的像素值获取预测直流分量，之后进行DLT变换，并结合索引残差值和DLT逆变换获取重建图像块，以便最终获得重建图像。因右下角像素点在33种角度模式中可更大的提高直流分量的值完全由有效像素值得出的概率，所以减少了SDC的编码过程中DLT产生的变换误差，能在VSO进行RDO选择时选择到更合适的编码结果，从而提升了图像处理结果的编码性能。
本发明的实施例提供一种视频编码装置，应用于多媒体领域，具体可以用于对深度图像的图像处理，可以应用于本发明的实施例提供的图像处理系统，可以结合本发明的上述实施例为图像处理系统中视频编码装置提供的图像处理方法进行使用，参照图12所示，该视频编码装置可以嵌入或本身就是微处理计算机，比如：通用计算机、客户定制机、手机终端或平板机等便携设备，该视频编码装置12001包括：至少一个数据端口12011、处理器12012、存储器12013和总线12014，该至少一个数据端口12011、处理器12012和存储器12013通过总线12014连接并完成相互间的通信。
该总线12014可以是ISA(Industry Standard Architecture，工业标准体系结构)总线、PCI(Peripheral Component，外部设备互连)总线或EISA(Extended Industry Standard Architecture，扩展工业标准体系结构)总线等。该总线12014可以分为地址总线、数据总线、控制总线等。为便于表示，图12中仅用一条粗线表示，但并不表示仅有一根总线或一种类型的总线。其中：
存储器12013用于存储可执行程序代码，该程序代码包括计算机操作指令。存储器12013可能包括高速RAM存储器，也可能还包括非易失性存储器(non-volatile memory)，例如至少一个磁盘存储器。
处理器12012可能是一个中央处理器(Central Processing Unit，简称为CPU)，或者是特定集成电路(Application Specific Integrated Circuit，简称为ASIC)，或者是被配置成实施本发明实施例的一个或多个集成电路。
其中，所述数据端口12011，用于获取待编码深度图像的原始图像块、至少一种图像处理模式和深度查询表，其中，所述深度查询表包含至少一个索引值以及每个索引值对应的像素值，所述索引值用于表示所述索引值自身对应的像素值；
所述数据端口12011还用于根据所述数据端口12011获取的所述至少一种图像处理模式，获取每种图像处理模式各自对应的预测图像块；
所述处理器12012，用于根据所述数据端口12011获取的所述预测图像块中一个像素点的像素值获取预测直流分量；
所述处理器12012，用于根据所述数据端口12011获取的所述深度查询表和所述直流分量单元获取的所述预测直流分量获得预测索引值。
所述处理器12012，用于根据所述数据端口12011获取的所述原始图像块获取原始直流分量；
所述处理器12012，用于根据所述数据端口12011获取的所述深度查询表和所述处理器12012获取的所述原始直流分量获得原始索引值；
所述处理器12012，用于将所述处理器12012获取的所述原始索引值减所述处理器12012获取的所述预测索引值得到索引残差值；
所述处理器12012，用于根据所述数据端口12011获取的所述预测图像块和预设规则从所述至少一种图像处理模式中选取优选的图像处理模式；
所述处理器12012，用于对所述处理器12012选取的所述优选的图像处理模式、所述处理器12012获取的所述优选的图像处理模式对应的索引残差值和所述数据端口12011获取的所述深度查询表进行编码以得到编码数据。
所述视频编码装置还可以向外发送所述编码数据，以便其他装置根据所述编码数据生成重建深度图像。具体可以由所述数据端口12011执行发送所述编码数据的功能。
可选的，所述处理器12012具体用于根据所述数据端口12011获取的所述预测图像块中右下角像素点的像素值获取预测直流分量；
或者，所述处理器12012具体用于根据所述数据端口12011获取的所述预测图像块中心区域的一个像素点的像素值获取预测直流分量，所述预测图像块中心区域是所述预测图像块中心处边长为2个像素的正方形图像区域。
可选的，所述处理器12012，还用于在所述数据端口12011获取的所述图像处理模式为角度模式时，根据所述数据端口12011获取的所述预测图像块中一个像素点的像素值获取预测直流分量。
可选的，所述处理器12012，还用于在所述数据端口12011获取的所述图像处理模式为平面模式时，根据所述数据端口12011获取的所述预测图像块中一个像素点的像素值获取预测直流分量。
可选的，所述处理器12012，还用于在所述数据端口12011获取的所述图像处理模式为直流分量模式时，根据所述数据端口12011获取的 所述预测图像块中一个像素点的像素值获取预测直流分量。
可选的，所述处理器12012具体用于根据所述数据端口12011获取的所述原始图像块中一个像素点的像素值获取原始直流分量。
可选的，所述处理器12012具体用于根据所述数据端口12011获取的所述原始图像块中右下角像素点的像素值获取原始直流分量；
或者，所述处理器12012具体用于根据所述数据端口12011获取的所述原始图像块中心区域的一个像素点的像素值获取原始直流分量，所述原始图像块中心区域是所述原始图像块中心处边长为2个像素的正方形图像区域。
可选的，所述数据端口12011具体用于在关闭帧内参考像素滤波器时，根据所述数据端口12011获取的所述至少一种图像处理模式，获取每种图像处理模式各自对应的预测图像块。
可选的，所述处理器12012，具体用于根据所述数据端口12011获取的所述预测图像块计算所述至少一种图像处理模式中与所述预测图像块对应的图像处理模式的率失真代价；
所述处理器12012，用于根据所述处理器12012计算的所述率失真代价，将所述数据端口12011获取的所述至少一种图像处理模式中率失真代价最小的图像处理模式作为优选的图像处理模式。
本发明实施例提供的视频编码装置，能获取待编码深度图像的原始图像块、至少一种图像处理模式和深度查询表，根据图像处理模式获取预测图像块，结合具体的图像处理模式根据所述预测图像块右下角像素点的像素值获取预测直流分量，并根据原始图像块右下角像素点的像素值得出原始直流分量，根据原始直流分量和预测直流分量进行DLT计算出索引残差值，以便根据索引残差值重建被编码的深度图像。因右下角像素点在33种角度模式中可更大的提高直流分量的值完全由有效像素值得出的概率，所以减少了SDC的编码过程中DLT产生的变换误差，能在VSO进行RDO选择时选择到更合适的编码结果，从而提升了图像处理结果的编码性能。
本发明的实施例提供一种视频解码装置，应用于多媒体领域，具体可以用于对深度图像的图像处理，可以应用于本发明的实施例提供的图像处理系统，可以结合本发明的上述实施例为图像处理系统中视频解码装置提供的图像处理方法进行使用，参照图13所示，该视频解码装置可以嵌入或本身就是微处理计算机，比如：通用计算机、客户定制机、手机终端或平板机等便携设备，该视频解码装置13001包括：至少一个数据端口13011、处理器13012、存储器13013和总线13014，该至少一个数据端口13011、处理器13012和存储器13013通过总线13014连接并完成相互间的通信。
该总线13014可以是ISA(Industry Standard Architecture，工 业标准体系结构)总线、PCI(Peripheral Component，外部设备互连)总线或EISA(Extended Industry Standard Architecture，扩展工业标准体系结构)总线等。该总线13014可以分为地址总线、数据总线、控制总线等。为便于表示，图13中仅用一条粗线表示，但并不表示仅有一根总线或一种类型的总线。其中：
存储器13013用于存储可执行程序代码，该程序代码包括计算机操作指令。存储器13013可能包括高速RAM存储器，也可能还包括非易失性存储器(non-volatile memory)，例如至少一个磁盘存储器。
处理器13012可能是一个中央处理器(Central Processing Unit，简称为CPU)，或者是特定集成电路(Application Specific Integrated Circuit，简称为ASIC)，或者是被配置成实施本发明实施例的一个或多个集成电路。
其中，所述数据端口13011，用于获取待重建深度图像的编码数据，所述编码数据包括图像处理模式、索引残差值和深度查询表，其中，所述深度查询表包含至少一个索引值以及每个索引值对应的像素值，所述索引值用于表示所述索引值自身对应的像素值；
所述数据端口13011还用于根据所述数据端口13011获取的所述图像处理模式，获取预测图像块；
所述处理器13012，用于根据所述数据端口13011获取的所述预测图像块中一个像素点的像素值获取预测直流分量；
所述处理器13012，用于根据所述数据端口13011获取的所述深度查询表和所述直流分量单元获取的所述预测直流分量获得预测索引值。
所述处理器13012，用于将所述数据端口13011获取的所述索引残差值加所述处理器13012获取的所述预测索引值得到重建索引值；
所述处理器13012，用于根据所述数据端口13011获取的所述深度查询表和所述处理器13012得到的所述重建索引值获得重建像素值，将所述重建像素值作为重建直流分量；
所述处理器13012，用于将所述处理器13012获得的所述重建直流分量减所述预测直流分量得到深度残差值；
所述处理器13012，用于为所述数据端口13011获取的所述预测图像块的每一个像素点的像素值加上所述处理器13012得到的所述深度残差值得到重建图像块，所述重建图像块用于重建所述待重建深度图像。
可选的，所述处理器13012具体用于根据所述数据端口13011获取的所述预测图像块中右下角像素点的像素值获取预测直流分量；
或者，所述处理器13012具体用于根据所述数据端口13011获取的所述预测图像块中心区域的一个像素点的像素值获取预测直流分量，所述预测图像块中心区域是所述预测图像块中心处边长为2个像素的 正方形图像区域。
可选的，所述处理器13012，还用于在所述数据端口13011获取的所述图像处理模式为角度模式时，根据所述数据端口13011获取的所述预测图像块中一个像素点的像素值获取预测直流分量。
可选的，所述处理器13012，还用于在所述数据端口13011获取的所述图像处理模式为平面模式时，根据所述数据端口13011获取的所述预测图像块中一个像素点的像素值获取预测直流分量。
可选的，所述处理器13012，还用于在所述数据端口13011获取的所述图像处理模式为直流分量模式时，根据所述数据端口13011获取的所述预测图像块中一个像素点的像素值获取预测直流分量。
可选的，所述数据端口13011具体用于在关闭帧内参考像素滤波器时，根据所述数据端口13011获取的所述图像处理模式，获取预测图像块。
本发明实施例提供的视频解码装置，能获取待重建深度图像的编码数据，所述编码数据包括图像处理模式、索引残差值和深度查询表，并结合具体的图像处理模式根据所述预测图像块中一个像素点的像素值获取预测直流分量，之后进行DLT变换，并结合索引残差值和DLT逆变换获取重建图像块，以便最终获得重建图像。因右下角像素点在33种角度模式中可更大的提高直流分量的值完全由有效像素值得出的概率，所以减少了SDC的编码过程中DLT产生的变换误差，能在VSO进行RDO选择时选择到更合适的编码结果，从而提升了图像处理结果的编码性能。
通过以上的实施方式的描述，所属领域的技术人员可以清楚地了解到本发明可以用硬件实现，或固件实现，或它们的组合方式来实现。当使用软件实现时，可以将上述功能存储在计算机可读介质中或作为计算机可读介质上的一个或多个指令或代码进行传输。计算机可读介质包括计算机存储介质和通信介质，其中通信介质包括便于从一个地方向另一个地方传送计算机程序的任何介质。存储介质可以是计算机能够存取的任何可用介质。以此为例但不限于：计算机可读介质可以包括RAM(Random Access Memory，随机存储器)、ROM(Read Only Memory，只读内存)、EEPROM(Electrically Erasable Programmable Read Only Memory，电可擦可编程只读存储器)、CD-ROM(Compact Disc Read Only Memory，即只读光盘)或其他光盘存储、磁盘存储介质或者其他磁存储设备、或者能够用于携带或存储具有指令或数据结构形式的期望的程序代码并能够由计算机存取的任何其他介质。此外，任何连接可以适当的成为计算机可读介质。例如，如果软件是使用同轴电缆、光纤光缆、双绞线、DSL(Digital Subscriber Line，数字用户专线)或者诸如红外线、无线电和微波之类的无线技术从网站、服务器或者其他远程源传输的，那么同轴电缆、光纤光缆、双绞线、DSL或者诸如红外线、无线和 微波之类的无线技术包括在所属介质的定影中。如本发明所使用的，盘和碟包括CD(Compact Disc，压缩光碟)、激光碟、光碟、DVD碟(Digital Versatile Disc，数字通用光)、软盘和蓝光光碟，其中盘通常磁性的复制数据，而碟则用激光来光学的复制数据。上面的组合也应当包括在计算机可读介质的保护范围之内。
本发明的各实施例提供的图像处理方法、装置及系统，可以扩展到深度图像的其他涉及DLT变换的技术中，例如基于DLT的DMM(Depth Modelling Mode，深度图像建模模式)模式的原始DC值的计算过程、基于DLT的帧间SDC的原始DC与预测DC的计算等等。
以上所述，仅为本发明的具体实施方式，但本发明的保护范围并不局限于此，任何熟悉本技术领域的技术人员在本发明揭露的技术范围内，可轻易想到变化或替换，都应涵盖在本发明的保护范围之内。因此，本发明的保护范围应所述以权利要求的保护范围为准。。

Claims (30)

  1. 一种图像处理方法，其特征在于，包括：
    获取待编码深度图像的原始图像块、至少一种图像处理模式和深度查询表，其中，所述深度查询表包含至少一个索引值以及每个索引值对应的像素值，所述索引值用于表示所述索引值自身对应的像素值；
    根据所述至少一种图像处理模式，获取每种图像处理模式各自对应的预测图像块；
    根据所述预测图像块中一个像素点的像素值获取预测直流分量；
    根据所述深度查询表和所述预测直流分量获得预测索引值；
    根据所述原始图像块获取原始直流分量；
    根据所述深度查询表和所述原始直流分量获得原始索引值；
    将所述原始索引值减所述预测索引值得到索引残差值；
    根据所述预测图像块和预设规则从所述至少一种图像处理模式中选取优选的图像处理模式；
    对所述优选的图像处理模式、所述优选的图像处理模式对应的索引残差值和所述深度查询表进行编码以得到编码数据。
  2. 根据权利要求1所述的方法，其特征在于，根据所述预测图像块中一个像素点的像素值获取预测直流分量，具体为：
    根据所述预测图像块中右下角像素点的像素值获取预测直流分量；
    或者，
    根据所述预测图像块中心区域的一个像素点的像素值获取预测直流分量，所述预测图像块中心区域是所述预测图像块中心处边长为2个像素的正方形图像区域。
  3. 根据权利要求1或2所述的方法，其特征在于，根据所述预测图像块中一个像素点的像素值获取预测直流分量，包括：
    在所述图像处理模式为角度模式时，根据所述预测图像块中一个像素点的像素值获取预测直流分量。
  4. 根据权利要求1-3任一项所述的方法，其特征在于，根据所述预测图像块中一个像素点的像素值获取预测直流分量，包括：
    在所述图像处理模式为平面模式时，根据所述预测图像块中一个像素点的像素值获取预测直流分量。
  5. 根据权利要求1-4任一项所述的方法，其特征在于，根据所述预测图像块中一个像素点的像素值获取预测直流分量，包括：
    在所述图像处理模式为直流分量模式时，根据所述预测图像块中一个像素点的像素值获取预测直流分量。
  6. 根据权利要求1-5任一项所述的方法，其特征在于，根据所述原始图像块获取原始直流分量，具体包括：
    根据所述原始图像块中一个像素点的像素值获取原始直流分量。
  7. 根据权利要求6所述的方法，其特征在于，根据所述原始图像 块中一个像素点的像素值获取原始直流分量，具体包括：
    根据所述原始图像块中右下角像素点的像素值获取原始直流分量；
    或者，
    根据所述原始图像块中心区域的一个像素点的像素值获取原始直流分量，所述原始图像块中心区域是所述原始图像块中心处边长为2个像素的正方形图像区域。
  8. 根据权利要求1-7任一项所述的方法，其特征在于，根据所述至少一种图像处理模式，获取每种图像处理模式各自对应的预测图像块，具体包括：
    在关闭帧内参考像素滤波器时，根据所述至少一种图像处理模式，获取每种图像处理模式各自对应的预测图像块。
  9. 根据权利要求1-8任一项所述的方法，其特征在于，根据所述预测图像块和预设规则从所述至少一种图像处理模式中选取优选的图像处理模式，具体包括：
    根据所述预测图像块计算所述至少一种图像处理模式中与所述预测图像块对应的图像处理模式的率失真代价；
    根据所述率失真代价，将所述至少一种图像处理模式中率失真代价最小的图像处理模式作为优选的图像处理模式。
  10. 一种图像处理方法，其特征在于，包括：
    获取待重建深度图像的编码数据，所述编码数据包括图像处理模式、索引残差值和深度查询表，其中，所述深度查询表包含至少一个索引值以及每个索引值对应的像素值，所述索引值用于表示所述索引值自身对应的像素值；
    根据所述图像处理模式，获取预测图像块；
    根据所述预测图像块中一个像素点的像素值获取预测直流分量；
    根据所述深度查询表和所述预测直流分量获得预测索引值；
    将所述索引残差值加所述预测索引值得到重建索引值；
    根据所述深度查询表和所述重建索引值获得重建像素值，将所述重建像素值作为重建直流分量；
    将所述重建直流分量减所述预测直流分量得到深度残差值；
    为所述预测图像块的每一个像素点的像素值加上所述深度残差值得到重建图像块，所述重建图像块用于重建所述待重建深度图像。
  11. 根据权利要求10所述的方法，其特征在于，根据所述预测图像块中一个像素点的像素值获取预测直流分量，具体为：
    根据所述预测图像块中右下角像素点的像素值获取预测直流分量；
    或者，
    根据所述预测图像块中心区域的一个像素点的像素值获取预测直流分量，所述预测图像块中心区域是所述预测图像块中心处边长为2个像素的正方形图像区域。
  12. 根据权利要求10或11所述的方法，其特征在于，根据所述预 测图像块中一个像素点的像素值获取预测直流分量，包括：
    在所述图像处理模式为角度模式时，根据所述预测图像块中一个像素点的像素值获取预测直流分量。
  13. 根据权利要求10-12任一项所述的方法，其特征在于，根据所述预测图像块中一个像素点的像素值获取预测直流分量，包括：
    在所述图像处理模式为平面模式时，根据所述预测图像块中一个像素点的像素值获取预测直流分量。
  14. 根据权利要求10-13任一项所述的方法，其特征在于，根据所述预测图像块中一个像素点的像素值获取预测直流分量，包括：
    在所述图像处理模式为直流分量模式时，根据所述预测图像块中一个像素点的像素值获取预测直流分量。
  15. 根据权利要求10-14任一项所述的方法，其特征在于，根据所述图像处理模式，获取预测图像块，具体包括：
    在关闭帧内参考像素滤波器时，根据所述图像处理模式，获取预测图像块。
  16. 一种视频编码装置，其特征在于，包括：
    获取单元，用于获取待编码深度图像的原始图像块、至少一种图像处理模式和深度查询表，其中，所述深度查询表包含至少一个索引值以及每个索引值对应的像素值，所述索引值用于表示所述索引值自身对应的像素值；
    所述获取单元还用于根据所述获取单元获取的所述至少一种图像处理模式，获取每种图像处理模式各自对应的预测图像块；
    预测直流分量单元，用于根据所述获取单元获取的所述预测图像块中一个像素点的像素值获取预测直流分量；
    预测索引值单元，用于根据所述获取单元获取的所述深度查询表和所述直流分量单元获取的所述预测直流分量获得预测索引值；
    原始直流分量单元，用于根据所述获取单元获取的所述原始图像块获取原始直流分量；
    原始索引值单元，用于根据所述获取单元获取的所述深度查询表和所述原始直流分量单元获取的所述原始直流分量获得原始索引值；
    索引残差值单元，用于将所述原始索引值单元获取的所述原始索引值减所述预测索引值单元获取的所述预测索引值得到索引残差值；
    优选单元，用于根据所述获取单元获取的所述预测图像块和预设规则从所述至少一种图像处理模式中选取优选的图像处理模式；
    编码单元，用于对所述优选单元选取的所述优选的图像处理模式、所述索引残差值单元获取的所述优选的图像处理模式对应的索引残差值和所述获取单元获取的所述深度查询表进行编码以得到编码数据。
  17. 根据权利要求16所述的装置，其特征在于，所述预测直流分量单元具体用于根据所述获取单元获取的所述预测图像块中右下角像素点的像素值获取预测直流分量；
    或者，
    所述预测直流分量单元具体用于根据所述获取单元获取的所述预测图像块中心区域的一个像素点的像素值获取预测直流分量，所述预测图像块中心区域是所述预测图像块中心处边长为2个像素的正方形图像区域。
  18. 根据权利要求16或17所述的装置，其特征在于，所述预测直流分量单元，还用于在所述获取单元获取的所述图像处理模式为角度模式时，根据所述获取单元获取的所述预测图像块中一个像素点的像素值获取预测直流分量。
  19. 根据权利要求16-18任一项所述的装置，其特征在于，所述预测直流分量单元，还用于在所述获取单元获取的所述图像处理模式为平面模式时，根据所述获取单元获取的所述预测图像块中一个像素点的像素值获取预测直流分量。
  20. 根据权利要求16-19任一项所述的装置，其特征在于，所述预测直流分量单元，还用于在所述获取单元获取的所述图像处理模式为直流分量模式时，根据所述获取单元获取的所述预测图像块中一个像素点的像素值获取预测直流分量。
  21. 根据权利要求16-20任一项所述的装置，其特征在于，所述原始直流分量单元具体用于根据所述获取单元获取的所述原始图像块中一个像素点的像素值获取原始直流分量。
  22. 根据权利要求21所述的装置，其特征在于，所述原始直流分量单元具体用于根据所述获取单元获取的所述原始图像块中右下角像素点的像素值获取原始直流分量；
    或者，
    所述原始直流分量单元具体用于根据所述获取单元获取的所述原始图像块中心区域的一个像素点的像素值获取原始直流分量，所述原始图像块中心区域是所述原始图像块中心处边长为2个像素的正方形图像区域。
  23. 根据权利要求16-22所述的装置，其特征在于，所述获取单元具体用于在关闭帧内参考像素滤波器时，根据所述获取单元获取的所述至少一种图像处理模式，获取每种图像处理模式各自对应的预测图像块。
  24. 根据权利要求16-23任一项所述的装置，其特征在于，所述优选单元，具体包括：
    率失真子单元，用于根据所述获取单元获取的所述预测图像块计算所述至少一种图像处理模式中与所述预测图像块对应的图像处理模式的率失真代价；
    优选子单元，用于根据所述率失真子单元计算的所述率失真代价，将所述获取单元获取的所述至少一种图像处理模式中率失真代价最小的图像处理模式作为优选的图像处理模式。
  25. 一种视频解码装置，其特征在于，包括：
    获取单元，用于获取待重建深度图像的编码数据，所述编码数据包括图像处理模式、索引残差值和深度查询表，其中，所述深度查询表包含至少一个索引值以及每个索引值对应的像素值，所述索引值用于表示所述索引值自身对应的像素值；
    所述获取单元还用于根据所述获取单元获取的所述图像处理模式，获取预测图像块；
    预测直流分量单元，用于根据所述获取单元获取的所述预测图像块中一个像素点的像素值获取预测直流分量；
    预测索引值单元，用于根据所述获取单元获取的所述深度查询表和所述直流分量单元获取的所述预测直流分量获得预测索引值；
    重建索引值单元，用于将所述获取单元获取的所述索引残差值加所述预测索引值单元获取的所述预测索引值得到重建索引值；
    重建直流分量单元，用于根据所述获取单元获取的所述深度查询表和所述重建索引值单元得到的所述重建索引值获得重建像素值，将所述重建像素值作为重建直流分量；
    深度残差值单元，用于将所述重建直流分量单元获得的所述重建直流分量减所述预测直流分量得到深度残差值；
    重建单元，用于为所述获取单元获取的所述预测图像块的每一个像素点的像素值加上所述深度残差值单元得到的所述深度残差值得到重建图像块，所述重建图像块用于重建所述待重建深度图像。
  26. 根据权利要求25所述的装置，其特征在于，所述预测直流分量单元具体用于根据所述获取单元获取的所述预测图像块中右下角像素点的像素值获取预测直流分量；
    或者，
    所述预测直流分量单元具体用于根据所述获取单元获取的所述预测图像块中心区域的一个像素点的像素值获取预测直流分量，所述预测图像块中心区域是所述预测图像块中心处边长为2个像素的正方形图像区域。
  27. 根据权利要求25或26所述的装置，其特征在于，所述预测直流分量单元，还用于在所述获取单元获取的所述图像处理模式为角度模式时，根据所述获取单元获取的所述预测图像块中一个像素点的像素值获取预测直流分量。
  28. 根据权利要求25-27任一项所述的装置，其特征在于，所述预测直流分量单元，还用于在所述获取单元获取的所述图像处理模式为平面模式时，根据所述获取单元获取的所述预测图像块中一个像素点的像素值获取预测直流分量。
  29. 根据权利要求25-28任一项所述的装置，其特征在于，所述预测直流分量单元，还用于在所述获取单元获取的所述图像处理模式为直流分量模式时，根据所述获取单元获取的所述预测图像块中一个像素点 的像素值获取预测直流分量。
  30. 根据权利要求16-29任一项所述的装置，其特征在于，所述获取单元具体用于在关闭帧内参考像素滤波器时，根据所述获取单元获取的所述图像处理模式，获取预测图像块。
PCT/CN2015/077721 2014-06-25 2015-04-28 一种图像处理方法、装置及系统 WO2015196860A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201410294340.XA CN104079942B (zh) 2014-06-25 2014-06-25 一种图像处理方法、装置及系统
CN201410294340.X 2014-06-25

Publications (1)

Publication Number Publication Date
WO2015196860A1 true WO2015196860A1 (zh) 2015-12-30

Family

ID=51600950

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2015/077721 WO2015196860A1 (zh) 2014-06-25 2015-04-28 一种图像处理方法、装置及系统

Country Status (2)

Country Link
CN (1) CN104079942B (zh)
WO (1) WO2015196860A1 (zh)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104079942B (zh) * 2014-06-25 2017-06-06 华为技术有限公司 一种图像处理方法、装置及系统
CN105637865B (zh) * 2015-02-05 2019-11-19 华为技术有限公司 图像预测处理方法和相关设备
WO2017132933A1 (zh) * 2016-02-04 2017-08-10 华为技术有限公司 一种图像处理方法以及相关装置

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20120183057A1 (en) * 2011-01-14 2012-07-19 Samsung Electronics Co., Ltd. System, apparatus, and method for encoding and decoding depth image
CN103237214A (zh) * 2013-04-12 2013-08-07 华为技术有限公司 深度图像的编解码方法和编解码装置
CN103813173A (zh) * 2014-03-14 2014-05-21 北方工业大学 基于bcim的3d-hevc快速帧内预测方案
CN104079942A (zh) * 2014-06-25 2014-10-01 华为技术有限公司 一种图像处理方法、装置及系统

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20120183057A1 (en) * 2011-01-14 2012-07-19 Samsung Electronics Co., Ltd. System, apparatus, and method for encoding and decoding depth image
CN103237214A (zh) * 2013-04-12 2013-08-07 华为技术有限公司 深度图像的编解码方法和编解码装置
CN103813173A (zh) * 2014-03-14 2014-05-21 北方工业大学 基于bcim的3d-hevc快速帧内预测方案
CN104079942A (zh) * 2014-06-25 2014-10-01 华为技术有限公司 一种图像处理方法、装置及系统

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
TIAN, TIAN ET AL.: "Next-Generation 3D Video Coding Technology Based on HEVC", PROCEEDINGS OF THE 15 TH NATIONAL CONSUMER ELECTRONIC TECHNOLOGY SEMINAR, 20 December 2013 (2013-12-20) *

Also Published As

Publication number Publication date
CN104079942A (zh) 2014-10-01
CN104079942B (zh) 2017-06-06

Similar Documents

Publication Publication Date Title
US20200260117A1 (en) Methods and Apparatuses for Coding and Decoding Depth Map
WO2019129130A1 (zh) 图像预测方法、装置以及编解码器
TW201919402A (zh) 視訊圖像編碼和解碼的方法、裝置及設備
KR102612489B1 (ko) 비디오 디코딩 방법 및 비디오 디코더
CN113923455B (zh) 一种双向帧间预测方法及装置
US20190289331A1 (en) Image processing apparatus for performing filtering on restored images and filtering method thereof
US9503751B2 (en) Method and apparatus for simplified depth coding with extended prediction modes
US20220191548A1 (en) Picture prediction method, encoder, decoder and storage medium
US20150365698A1 (en) Method and Apparatus for Prediction Value Derivation in Intra Coding
JP2022524916A (ja) 適応的な数の領域を伴う幾何学的分割のための形状適応離散コサイン変換
WO2015196860A1 (zh) 一种图像处理方法、装置及系统
WO2020140215A1 (zh) 色度帧内预测方法和装置、及计算机存储介质
KR102464520B1 (ko) 적응형 곱셈 계수를 이용한 이미지 필터링을 위한 방법 및 장치
JP6457248B2 (ja) 画像復号装置、画像符号化装置および画像復号方法
CN112470468A (zh) 解码预测方法、装置及计算机存储介质
AU2010202963A1 (en) Frame rate up-sampling for multi-view video coding using distributing video coding principles
JP6539580B2 (ja) インター予測装置、インター予測方法、動画像符号化装置、動画像復号装置及びコンピュータ可読記録媒体
KR102402671B1 (ko) 보간 필터의 연산 복잡도를 조절할 수 있는 영상 처리 장치, 영상 보간 방법 및 영상 부호화 방법
WO2013075515A1 (zh) 多视点视频的编码、解码方法、装置和编解码器
WO2020063687A1 (zh) 视频解码方法及视频解码器
KR20200004348A (ko) 타겟 영역 수정을 통해 비디오 신호를 처리하는 방법 및 장치
JP7494303B2 (ja) 符号化及び復号化方法並びに装置
JP2013157950A (ja) 符号化方法、復号方法、符号化装置、復号装置、符号化プログラム及び復号プログラム
US20170019683A1 (en) Video encoding apparatus and method and video decoding apparatus and method
WO2020048361A1 (zh) 视频解码方法及视频解码器

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 15812654

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 15812654

Country of ref document: EP

Kind code of ref document: A1