WO2016110249A1 - 处理图像的方法和装置 - Google Patents

处理图像的方法和装置 Download PDF

Info

Publication number
WO2016110249A1
WO2016110249A1 PCT/CN2016/070228 CN2016070228W WO2016110249A1 WO 2016110249 A1 WO2016110249 A1 WO 2016110249A1 CN 2016070228 W CN2016070228 W CN 2016070228W WO 2016110249 A1 WO2016110249 A1 WO 2016110249A1
Authority
WO
WIPO (PCT)
Prior art keywords
image
transform coefficients
transform
target
determining
Prior art date
Application number
PCT/CN2016/070228
Other languages
English (en)
French (fr)
Inventor
赵寅
杨海涛
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
Priority to EP16734909.1A priority Critical patent/EP3236658B1/en
Publication of WO2016110249A1 publication Critical patent/WO2016110249A1/zh
Priority to US15/645,052 priority patent/US10291935B2/en

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
    • H04N19/60—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding
    • 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
    • H04N19/60—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding
    • H04N19/61—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding in combination with predictive coding
    • 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
    • H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/134—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or criterion affecting or controlling the adaptive coding
    • H04N19/136—Incoming video signal characteristics or properties
    • 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
    • H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/169—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding
    • H04N19/18—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding the unit being a set of transform coefficients
    • 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
    • H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/169—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding
    • H04N19/182—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding the unit being a pixel
    • 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
    • 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
    • H04N19/80—Details of filtering operations specially adapted for video compression, e.g. for pixel interpolation

Definitions

  • the present invention relates to the field of domain image processing and, more particularly, to a method and apparatus for processing images.
  • Video coding such as H.264/AVC, H.265/HEVC, AVS and other video coding standards, usually adopt a hybrid coding framework, mainly including prediction, transform, quantization, entropy coding. ) and other links.
  • Video decoding is a process of converting a code stream into a video image, which includes entropy decoding, prediction, dequantization, and inverse transform. First, the code stream is parsed by the entropy decoding process to encode the mode information and the quantized transform coefficients.
  • the prediction pixel is obtained from the coding mode information and the reconstructed pixel that has been decoded; on the other hand, the quantized transform coefficient is inverse quantized to obtain the reconstructed transform coefficient, and then the reconstructed transform coefficient is inversely transformed to obtain a reconstructed Residual information. Thereafter, the reconstructed residual information is added to the predicted pixels to obtain reconstructed pixels, thereby recovering the video image.
  • reconstructed pixels may not be identical to original pixels, and the numerical difference between the two is called distortion.
  • distortion is caused by quantization.
  • QP quantization parameter
  • the pixel quality is worse.
  • the knowledge base based encoding is an extension of H.264/AVC, H.265/HEVC.
  • the decoding end contains a knowledge base in which some images (images) and/or image regions are stored. Call it a patch.
  • the image or pattern in the knowledge base may be from the reconstructed image that has been decoded in the currently decoded video. For example, some representative images are extracted from the decoded reconstructed image and added to the knowledge base; the image or pattern in the knowledge base may also be from the currently decoded video.
  • these pre-stored images or patterns may be original images that have not been encoded and compressed.
  • the predicted pixel information utilized in decoding may be derived from pixel information in the knowledge base.
  • the prediction link uses the reconstructed pixels of the coded region to generate predicted pixels of the original pixels corresponding to the current coded block.
  • the prediction mode mainly includes two categories: intra prediction and inter prediction. Among them, the template matching in the intra coding and the decoder side motion vector derivation in the inter coding need to use the reconstructed image template area around the current decoding prediction block, at present One or more nearest neighbor images in the reconstructed region or other reconstructed frames in the frame that have the smallest difference from the current decoded block template region are referred to as matching images. For these two types of techniques, how to evaluate the image difference value of a candidate template area image in the template area image and the matching process or the distance in the range space is a key issue, which directly determines the final search. result.
  • Image difference value calculation methods between two conventional images such as Sum of Squared Errors (SSE), Sum of Absolute Difference (SAD), Mean Squared Error (Mean Squared Error) (MSE), Mean Absolute Difference (MAD), and, for example, the Sum of Absolute Transformed Difference (SATD) of the transform coefficient domain of the two images after Hadamard transform.
  • SSE Sum of Squared Errors
  • SAD Sum of Absolute Difference
  • MSE Mean Squared Error
  • MAD Mean Absolute Difference
  • SSE Sum of Absolute Transformed Difference
  • Embodiments of the present invention provide a method and apparatus for processing an image to improve the accuracy of image processing.
  • a method of processing an image comprising: transforming a first image to obtain a first set of transform coefficients; performing the transform on the second image, or a difference image of the first image and the second image Performing the transform to obtain a second set of transform coefficients; selecting, according to the magnitude of the transform coefficients in the first set of transform coefficients, a first set of transform coefficients from the first set of transform coefficients, wherein the first set of transform coefficients The amplitude of the transform coefficient in the image satisfies a preset threshold condition; determining, according to the first set of transform coefficients and the second set of transform coefficients, an image difference value between the first image and the second image; a value that processes the first image and the second image.
  • the second set of transform coefficients is a set of transform coefficients obtained by performing the transform on the second image, according to the first set of transform coefficients And determining, by the second set of transform coefficients, the image difference value between the first image and the second image, comprising: according to a one-to-one correspondence between the first transform coefficient set and the second transform coefficient set transform coefficient, And selecting, from the second set of transform coefficients, a second set of transform coefficients corresponding to the first set of transform coefficients; and determining the image difference value according to the first set of transform coefficients and the second set of transform coefficients.
  • the determining the image difference according to the first set of transform coefficients and the second set of transform coefficients The value includes: determining a difference between each of the first set of transform coefficients and each of the second set of transform coefficients; and determining a sum of difference values of the respective transform coefficients as the image difference value.
  • the first set of transform coefficients and the second set of transform coefficients are all divided into N subgroups in the same manner. Determining the image difference value according to the first set of transform coefficients and the second set of transform coefficients, comprising: determining N target values, wherein an ith target value of the N target values And a sum of differences between the i-th sub-group of the first set of transform coefficients and the corresponding transform coefficients of the i-th sub-group of the second set of transform coefficients; according to the weighting coefficients of the N sub-groups, The N target values are weighted and summed to obtain the image difference value.
  • the determining the image difference according to the first set of transform coefficients and the second set of transform coefficients a value comprising: determining a first target value, wherein the first target value is a sum of differences between the first set of transform coefficients and respective ones of the second set of transform coefficients; determining a second target value And the second target value is a residual transform coefficient of the first transform coefficient set except the first set of transform coefficients and a residual transform of the second transform coefficient set except the second set of transform coefficients a sum of differences of the corresponding transform coefficients in the coefficients; weighting and summing the first target value and the second target value according to the weighting coefficients of the first target value and the second target value, The image difference value.
  • the second set of transform coefficients is a difference image of the first image and the second image Performing the transformed transform coefficient set, and determining, according to the first set of transform coefficients and the second transform coefficient set, an image difference value between the first image and the second image, including: according to the first a one-to-one correspondence between a transform coefficient set and the second transform coefficient set transform coefficient, and selecting, from the second transform coefficient set, a second set of transform coefficients corresponding to the first set of transform coefficients; A set of transform coefficients to determine the image difference value.
  • the determining, according to the second set of transform coefficients, the image difference value, The sum of the transform coefficients of the two sets of transform coefficients is determined as the image difference value.
  • the second set of transform coefficients includes N subgroups
  • the determining, according to the second set of transform coefficients, The image difference value includes: determining N target values, wherein the ith target value is a sum of transform coefficients in the i-th sub-group; and the N targets are based on the weighting coefficients of the N target values The values are weighted and summed to obtain the image difference value.
  • the determining, according to the second set of transform coefficients, the image difference value comprises: determining a first target a value, the first target value is a sum of transform coefficients of the second set of transform coefficients; determining a second target value, wherein the second target value is the second transform coefficient set except the second set of transforms a sum of residual transform coefficients outside the coefficients; weighting and summing the first target value and the second target value according to the weighting coefficients of the first target value and the second target value, to obtain the image Difference value.
  • the first image is a target image
  • the second image is any candidate image of the K candidate images.
  • processing, according to the image difference value, the first image and the second image including: determining, according to the target image and an image difference value of each candidate image, each of the candidate images a weighting coefficient; determining a filtered image of the target image according to pixel values of the K candidate images and weighting coefficients of the K candidate images.
  • the determining the candidate image according to the image difference value of the target image and each candidate image Weighting factor including: Determining weighting coefficients of each of the candidate images, wherein b1, a1, and h1 each represent a positive real number, Dk represents an image difference value of the first image and the kth candidate image, and Wk represents the kth candidate image Weighting factor.
  • Wk represents the weighting coefficient of the kth candidate image.
  • the determining the candidate image according to the image difference value of the target image and each candidate image Corresponding weighting factors, including: Determining weighting coefficients of each of the candidate images, wherein b3, a3, and h3 each represent a positive real number, Dk represents an image difference value of the first image and the kth candidate image, and Wk represents the kth candidate image Weighting factor.
  • the pixel values according to the K candidate images, and the weighting coefficients of the K candidate images Determining a filtered image of the target image, comprising: Determining the filtered image, wherein Pf j represents a pixel value of the filtered image at a jth pixel point, W0 represents a weighting coefficient of the first image, Wk represents a weighting coefficient of the kth candidate image, and P0 j represents The pixel value of the first image at the jth pixel point, and Pk j represents the pixel value of the kth candidate image at the jth pixel point.
  • a second aspect provides an apparatus for processing an image, including: a first transform unit configured to transform a first image to obtain a first transform coefficient set; and a second transform unit configured to perform the transform on the second image Or performing the transform on the difference image of the first image and the second image to obtain a second transform coefficient set; and selecting a unit, where the transform coefficient of the first transform coefficient set obtained according to the first transform unit is used The amplitude of the first transform coefficient is selected from the first set of transform coefficients, wherein an amplitude of the transform coefficients in the first set of transform coefficients satisfies a preset threshold condition; and a determining unit is configured to perform, according to the selecting unit Determining, by the selected first set of transform coefficients and the second set of transform coefficients obtained by the second transform unit, an image difference value between the first image and the second image; and a processing unit, configured to determine according to the determining The image difference value determined by the unit processes the first image and the second image.
  • the second transform coefficient set is a transform coefficient set obtained by performing the transform on the second image, where the determining unit is specifically configured to be used according to the a one-to-one correspondence between the first transform coefficient set and the second transform coefficient set transform coefficient, and selecting a second set of transform systems corresponding to the first set of transform coefficients from the second transform coefficient set And determining the image difference value according to the first set of transform coefficients and the second set of transform coefficients.
  • the determining unit is specifically configured to determine the first group of transform coefficients and the second group of transform coefficients a difference between each corresponding transform coefficient; determining a sum of differences of the respective transform coefficients as the image difference value.
  • the first set of transform coefficients and the second set of transform coefficients are all divided into N subgroups in the same manner.
  • the determining unit is specifically configured to determine N target values, where an ith target value of the N target values is an ith subgroup and a second set of transform coefficients in the first set of transform coefficients And a sum of differences of the corresponding transform coefficients in the i-th subgroup; and weighting the N target values according to the weighting coefficients of the N subgroups to obtain the image difference value.
  • the determining unit is specifically configured to determine a first target value, where the first target value is the a sum of a difference between a first set of transform coefficients and respective ones of the second set of transform coefficients; determining a second target value, wherein the second target value is the first transform coefficient set except the first a sum of a residual transform coefficient outside a set of transform coefficients and a difference between each of the remaining transform coefficients in the second transform coefficient set except the second set of transform coefficients; according to the first target value and And a weighting coefficient of the second target value, and weighting and summing the first target value and the second target value to obtain the image difference value.
  • the second set of transform coefficients is a difference image of the first image and the second image
  • the determining unit is specifically configured to select, according to the one-to-one correspondence between the first transform coefficient set and the second transform coefficient set transform coefficient, from the second transform coefficient set a second set of transform coefficients corresponding to the first set of transform coefficients; determining the image difference value based on the second set of transform coefficients.
  • the determining unit is specifically configured to determine a sum of each of the transform coefficients in the second set of transform coefficients as the image difference value.
  • the second set of transform coefficients includes N subgroups, and the determining unit is specifically configured to determine N target values.
  • the ith target value is a sum of transform coefficients in the i-th sub-group; and the N target values are weighted and summed according to the weighting coefficients of the N target values to obtain the image difference value.
  • the determining unit is specifically configured to determine a first target value, where the first target value is the second a sum of transform coefficients of the group of transform coefficients; determining a second target value, wherein the second target value is a sum of the remaining transform coefficients of the second transform coefficient set other than the second set of transform coefficients; And a weighting coefficient of the target value and the second target value, and weighting the first target value and the second target value to obtain the image difference value.
  • the first image is a target image
  • the second image is any candidate image of the K candidate images.
  • the processing unit is specifically configured to determine a weighting coefficient of each candidate image according to the target image and an image difference value of each candidate image; according to the pixel values of the K candidate images, and the K A weighting coefficient of the candidate image determines a filtered image of the target image.
  • the determining unit is specifically configured to be used according to Determining weighting coefficients of each of the candidate images, wherein b1, a1, and h1 each represent a positive real number, Dk represents an image difference value of the first image and the kth candidate image, and Wk represents the kth candidate image Weighting factor.
  • the weighting coefficient of the image wherein b2, a2, and h2 each represent a positive real number, Dk represents an image difference value of the first image and the kth candidate image, and Wk represents a weighting coefficient of the kth candidate image.
  • the determining unit is specifically configured to Determining weighting coefficients of each of the candidate images, wherein b3, a3, and h3 each represent a positive real number, Dk represents an image difference value of the first image and the kth candidate image, and Wk represents the kth candidate image Weighting factor.
  • the processing unit is specifically configured to Determining the filtered image, wherein Pf j represents a pixel value of the filtered image at a jth pixel point, W0 represents a weighting coefficient of the first image, Wk represents a weighting coefficient of the kth candidate image, and P0 j represents The pixel value of the first image at the jth pixel point, and Pk j represents the pixel value of the kth candidate image at the jth pixel point.
  • a decoding method including: selecting N candidate images from a knowledge base, wherein the N candidate images are used to determine pixel prediction values of an image to be decoded, and each candidate image and the candidate image The shape and the size of the decoded image are the same, the image to be decoded includes at least one image block to be decoded, and the same transformation is performed on each of the image to be decoded and each of the N target images to obtain a transformation of the image to be decoded.
  • the transform coefficient set of the image to be decoded is in one-to-one correspondence with the transform coefficient of the transform coefficient set of each target image; and the amplitude of the transform coefficient set from the image to be decoded is selected to satisfy Presetting a transform coefficient of the threshold to obtain a first set of transform coefficients; and selecting, according to the first set of transform coefficients, a set of transform coefficients from each of the target images a second set of transform coefficients corresponding to the set of transform coefficients; determining, according to the second set of transform coefficients corresponding to the N target images, image difference values of the image to be decoded and the N candidate images; And determining, by the image difference value of the image and the N candidate images, a pixel prediction value of the image to be de
  • the N target images are Determining image difference values of the image to be decoded and the N candidate images according to the second set of transform coefficients corresponding to the N target images, including: according to the first group of transforms The coefficient and the second set of transform coefficients corresponding to each candidate image determine an image difference value of the image to be decoded and each of the candidate images.
  • the determining, according to the first set of transform coefficients and a second set of transform coefficients corresponding to each candidate image Determining, by the image to be decoded, an image difference value of each of the candidate images, determining a difference between each of the first set of transform coefficients and each of the second set of transform coefficients; The difference of the transform coefficients is summed to obtain the image difference value.
  • the determining, according to the first set of transform coefficients and a second set of transform coefficients corresponding to each candidate image, And the image difference value of the image to be decoded and each of the candidate images includes: determining N target values, wherein the first group of transform coefficients and the second set of transform coefficients are divided into N subgroups in the same manner
  • the i-th target value of the N target values is a difference between the i-th sub-group of the first set of transform coefficients and the corresponding transform coefficient of the i-th sub-group of the second set of transform coefficients
  • the determining, according to the first set of transform coefficients and a second set of transform coefficients corresponding to each candidate image, The image difference value of the image to be decoded and each of the candidate images includes: determining a first target value, wherein the first target value is the first set of transform coefficients and the second set of transform coefficients a sum of difference values of respective corresponding transform coefficients; determining a second target value, wherein the second target value is a residual transform coefficient of the transform coefficient set of the image to be decoded other than the first set of transform coefficients and a transform coefficient of each candidate image, a sum of differences of respective transform coefficients of the remaining transform coefficients except the second set of transform coefficients; and a weighting coefficient according to the first target value and the second target value, Weighting and summing the first target value and the second target value to obtain the Image difference value.
  • the N target images are the N candidate images and the to-be-decoded image respectively at corresponding pixel points
  • An image obtained by the difference, the image difference value of the image to be decoded and the N candidate images is determined according to the second set of transform coefficients corresponding to the N target images, including: according to the N target images And determining, by the second set of transform coefficients corresponding to each of the target images, an image difference value of the candidate image corresponding to the target image, wherein the candidate image corresponding to each target image is The image to be decoded is obtained by subtracting the corresponding pixel points to obtain each of the target images.
  • the determining, according to the second group of transform coefficients corresponding to each target image in the N target images, And determining, by the decoded image, the image difference value of the candidate image corresponding to each of the target images comprising: determining a first target value, where the first target value is a sum of significant transform coefficients of the second set of transform coefficients; Determining a second target value, the second target value being a sum of the remaining transform coefficients of the transform coefficient set of each image except the second set of transform coefficients; according to the first target value and And a weighting coefficient of the second target value, and weighting and summing the first target value and the second target value to obtain the image difference value.
  • the determining, according to the image difference value of the image to be decoded and the plurality of candidate images, determining the to-be-determined Decoding the pixel prediction value of the image comprising: selecting one candidate image having the smallest image difference value from the plurality of candidate images according to the image difference value of the image to be decoded and the plurality of candidate images; The pixel value of the candidate image having the smallest value is determined as the pixel predicted value of the image to be decoded.
  • the determining, according to the image difference value of the image to be decoded and the plurality of candidate images, determining the to-be-determined Decoding the pixel prediction value of the image comprising: selecting E candidate images with the smallest image difference value from the plurality of candidate images according to the image difference value of the image to be decoded and the plurality of candidate images, E ⁇ 2 Determining a weight of each of the candidate images according to an image difference value of the image to be decoded and each of the E candidate images; and selecting the E candidate image according to weights of the E candidate images
  • the pixel values are weighted and averaged to obtain a pixel prediction value of the image to be decoded.
  • the selecting the N candidate images from the knowledge base includes: acquiring the N candidates from the code stream. An index of the image in the knowledge base; the N candidate images are selected from the knowledge base according to the index.
  • an encoding method including: selecting N candidate images from a knowledge base, wherein the N candidate images are used to determine pixel prediction values of an image to be encoded, and each candidate image and the candidate image The shape and size of the encoded image are the same, the image to be encoded includes at least one image block to be encoded; and the same transformation is performed on each of the image to be encoded and each of the N target images to obtain a transformation of the image to be encoded.
  • the transform coefficient set of the code image is in one-to-one correspondence with the transform coefficients in the transform coefficient set of each target image; and the transform coefficients satisfying the preset threshold are selected from the transform coefficient set of the image to be encoded to obtain the first set of transforms.
  • a coefficient according to the first set of transform coefficients selecting a second set of transform coefficients corresponding to the first set of transform coefficients from the transform coefficient set of each target image; and corresponding to the N target images Determining, by the two sets of transform coefficients, an image difference value of the image to be encoded and the N candidate images; determining a pixel prediction of the image to be encoded according to the image difference value of the image to be encoded and the N candidate images a value; encoding the image to be encoded according to the pixel prediction value.
  • the N target images are the N candidate images
  • the determining, according to the second group of transform coefficients corresponding to the N target images, The image difference value of the image to be encoded and the N candidate images includes: determining the image to be encoded and each candidate according to the first set of transform coefficients and a second set of transform coefficients corresponding to each candidate image The image difference value of the image.
  • the determining, according to the first set of transform coefficients and the second set of transform coefficients corresponding to each candidate image Determining, by the image to be encoded, an image difference value of each of the candidate images, determining a difference between each of the first set of transform coefficients and each of the second set of transform coefficients; The difference of the transform coefficients is summed to obtain the image difference value.
  • the determining, according to the first set of transform coefficients and the second set of transform coefficients corresponding to each candidate image, And the image difference value of the image to be encoded and the candidate image includes: determining N target values, wherein the first group of transform coefficients and the second set of transform coefficients are divided into N subgroups in the same manner
  • the i-th target value of the N target values is a difference between the i-th sub-group of the first set of transform coefficients and the corresponding transform coefficient of the i-th sub-group of the second set of transform coefficients
  • the determining, according to the first set of transform coefficients and the second set of transform coefficients corresponding to each candidate image, The image difference value of the image to be encoded and each of the candidate images includes: determining a first target value, wherein the first target value is the first set of transform coefficients and the second set of transform coefficients a sum of difference values of respective corresponding transform coefficients; determining a second target value, wherein the second target value is a residual transform coefficient of the transform coefficient set of the image to be encoded other than the first set of transform coefficients and a transform coefficient of each candidate image, a sum of differences of respective transform coefficients of the remaining transform coefficients except the second set of transform coefficients; and a weighting coefficient according to the first target value and the second target value, And weighting the first target value and the second target value to obtain the image difference value.
  • the N target images are the N candidate images and the to-be-coded image respectively at corresponding pixel points
  • An image obtained by the difference, the image difference value of the image to be encoded and the N candidate images is determined according to the second set of transform coefficients corresponding to the N target images, including: according to the N target images And determining, by the second set of transform coefficients corresponding to each of the target images, an image difference value of the candidate image corresponding to the target image, wherein the candidate image corresponding to each target image is The image to be encoded is differentiated at corresponding pixel points to obtain each of the target images.
  • the determining, according to the second group of transform coefficients corresponding to each target image in the N target images, The image difference value of the candidate image corresponding to the coded image and the target image includes: determining a first target value, where the first target value is a sum of significant transform coefficients of the second set of transform coefficients; Determining a second target value, the second target value being a sum of remaining transform coefficients of the transform coefficient set of each image other than the second set of transform coefficients; according to the first target value and the second a weighting coefficient of the target value, weighting and summing the first target value and the second target value to obtain the image difference value.
  • the determining, according to the image difference value of the image to be encoded and the plurality of candidate images, determining the to The pixel prediction value of the encoded image includes: selecting one candidate image having the smallest image difference value from the plurality of candidate images according to the image difference value of the image to be encoded and the plurality of candidate images; The pixel value of the candidate image having the smallest value is determined as the pixel predicted value of the image to be encoded.
  • the determining, according to the image difference value of the image to be encoded and the plurality of candidate images, determining the to The pixel prediction value of the encoded image includes: selecting E candidate images having the smallest image difference value from the plurality of candidate images according to the image difference value of the image to be encoded and the plurality of candidate images, E ⁇ 2 Determining a weight of each of the candidate images according to an image difference value of the image to be encoded and each of the E candidate images; and selecting the E candidate image according to weights of the E candidate images The pixel values are weighted and averaged to obtain a pixel prediction value of the image to be encoded.
  • the method further includes: writing an index of the N candidate images in the knowledge base into a code flow.
  • a decoder including:
  • a first selecting unit configured to select N candidate images from the knowledge base, wherein the N candidate images are used to determine pixel prediction values of the image to be decoded, and each candidate image and the shape of the image to be decoded The sizes are all the same, and the image to be decoded includes at least one image block to be decoded;
  • a transform unit configured to perform the same transform on each of the image to be decoded and each of the N target images, to obtain a transform coefficient set of the image to be decoded and a transform coefficient set of each target image, the N
  • the target image is the N candidate images, or the N target images are images obtained by respectively obtaining the difference between the N candidate images and the image to be decoded at a corresponding pixel, and the transform coefficient set of the image to be decoded One-to-one correspondence with transform coefficients in the transform coefficient set of each target image;
  • a second selecting unit configured to select, from the transform coefficient set of the image to be decoded, a transform coefficient whose amplitude satisfies a preset threshold, to obtain a first set of transform coefficients
  • a third selecting unit configured to select, according to the first set of transform coefficients, a second set of transform coefficients corresponding to the first set of transform coefficients from the transform coefficient set of each target image;
  • a first determining unit configured to determine, according to the second set of transform coefficients corresponding to the N target images, an image difference value between the image to be decoded and the N candidate images
  • a second determining unit configured to determine, according to the image difference value of the image to be decoded and the N candidate images, a pixel prediction value of the image to be decoded
  • a decoding unit configured to decode the to-be-decoded image according to the pixel prediction value.
  • the N target images are the N candidate images
  • the first determining unit is specifically configured to use the first group of transform coefficients and each And determining, by the second set of transform coefficients corresponding to the candidate image, an image difference value of the image to be decoded and each of the candidate images.
  • the first determining unit is specifically configured to determine the first group of transform coefficients and the second group of changes And a difference value of each corresponding transform coefficient in the coefficient; summing the difference values of the corresponding transform coefficients to obtain the image difference value.
  • the first determining unit is specifically configured to determine N target values, where the first group of transform coefficients And dividing the second set of transform coefficients into N subgroups in the same manner, wherein an ith target value of the N target values is an ith subgroup and a second set of transforms in the first set of transform coefficients And a sum of differences of the corresponding transform coefficients in the i-th sub-group of the coefficients; and weighting the N target values according to respective weighting coefficients of the N target values to obtain the image difference value.
  • the first determining unit is specifically configured to determine a first target value, where the first target value is a sum of a difference between the first set of transform coefficients and each of the second transform coefficients; a second target value, wherein the second target value is a transform coefficient set of the image to be decoded a sum of a difference of the remaining transform coefficients other than the first set of transform coefficients and a difference of each of the remaining transform coefficients of the transform coefficients of each of the candidate images other than the second set of transform coefficients; a weighting coefficient of the first target value and the second target value, and weighting the first target value and the second target value to obtain the image difference value.
  • the N target images are the N candidate images and the to-be-decoded image respectively at corresponding pixel points Determining the obtained image, the first determining unit is specifically configured to determine, according to the second set of transform coefficients corresponding to each of the N target images, the image to be decoded corresponding to each target image An image difference value of the candidate image, wherein the candidate image corresponding to each target image and the image to be decoded are deviated at corresponding pixel points to obtain each of the target images.
  • the first determining unit is specifically configured to determine a sum of each transform coefficient in the second set of transform coefficients The difference value for the image.
  • the first determining unit is specifically configured to determine N target values, wherein the second set of transform coefficients is divided into N subgroups, and an ith target value of the N target values is an i th sub a sum of transform coefficients in the group; and weighting the N target values according to weighting coefficients of the N target values to obtain the image difference value.
  • the first determining unit is specifically configured to determine a first target value, where the first target value is a sum of significant transform coefficients of the second set of transform coefficients; determining a second target value, wherein the second target value is a sum of residual transform coefficients of the transform coefficient set of each image other than the second set of transform coefficients And weighting the first target value and the second target value according to the weighting coefficients of the first target value and the second target value to obtain the image difference value.
  • the second determining unit is specifically configured to use the image to be decoded and the image of the multiple candidate images a difference value, one candidate image having the smallest image difference value is selected from the plurality of candidate images; and a pixel value of the candidate image having the smallest difference value is determined as a pixel predicted value of the image to be decoded.
  • the second determining unit is specifically configured to use the image to be decoded and the image of the multiple candidate images a difference value, the E candidate images having the smallest image difference value are selected from the plurality of candidate images, E ⁇ 2; according to the image difference value of the image to be decoded and each of the E candidate images, Determining a weight of each candidate image; performing weighted averaging on pixel values of the E candidate image according to weights of the E candidate images to obtain a pixel prediction value of the image to be decoded.
  • the first selecting unit is specifically configured to acquire the N candidate images from the code stream in the knowledge An index in the library; the N candidate images are selected from the knowledge base according to the index.
  • an encoder including:
  • a first selecting unit configured to select N candidate images from the knowledge base, wherein the N The candidate image is used to determine a pixel prediction value of the image to be encoded, and each candidate image is identical in shape and size to the image to be encoded, and the image to be encoded includes at least one image block to be encoded;
  • a transform unit configured to perform the same transform on each of the image to be encoded and each of the N target images, to obtain a transform coefficient set of the image to be encoded and a transform coefficient set of each target image, the N
  • the target image is the N candidate images, or the N target images are images obtained by respectively obtaining the difference between the N candidate images and the image to be encoded at a corresponding pixel, and the transform coefficient set of the image to be encoded One-to-one correspondence with transform coefficients in the transform coefficient set of each target image;
  • a second selecting unit configured to select, from the transform coefficient set of the image to be encoded, a transform coefficient whose amplitude satisfies a preset threshold, to obtain a first set of transform coefficients
  • a third selecting unit configured to select, according to the first set of transform coefficients, a second set of transform coefficients corresponding to the first set of transform coefficients from the transform coefficient set of each target image;
  • a first determining unit configured to determine, according to the second set of transform coefficients corresponding to the N target images, an image difference value between the image to be encoded and the N candidate images
  • a second determining unit configured to determine, according to the image difference value of the image to be encoded and the N candidate images, a pixel prediction value of the image to be encoded
  • a coding unit configured to encode the image to be encoded according to the pixel prediction value.
  • the N target images are the N candidate images
  • the first determining unit is specifically configured to use the first group of transform coefficients and each And determining, by the second set of transform coefficients corresponding to the candidate image, an image difference value of the image to be encoded and each of the candidate images.
  • the first determining unit is specifically configured to determine the first group of transform coefficients and the second group of transforms a difference between each corresponding transform coefficient in the coefficient; summing the differences of the respective transform coefficients to obtain the image difference value.
  • the first determining unit is specifically configured to determine N target values, wherein the first set of transform coefficients and the second set of transform coefficients are divided into N subgroups in the same manner, the N target values
  • the i-th target value is a sum of differences of the corresponding transform coefficients of the i-th sub-group of the first set of transform coefficients and the i-th sub-group of the second set of transform coefficients;
  • Each of the target values has a weighting coefficient, and the N target values are weighted and summed to obtain the image difference value.
  • the first determining unit is specifically configured to determine a first target value, where the first target value is a sum of a difference between the first set of transform coefficients and each of the second transform coefficients; a second target value, wherein the second target value is a transform coefficient set of the image to be encoded a sum of a difference of the remaining transform coefficients other than the first set of transform coefficients and a difference of each of the remaining transform coefficients of the transform coefficients of each of the candidate images other than the second set of transform coefficients; a weighting coefficient of the first target value and the second target value, and weighting the first target value and the second target value to obtain the image difference value.
  • the N target images are the N candidate images and the to-be-coded image respectively at corresponding pixel points
  • the first determining unit is specifically configured to determine, according to the second set of transform coefficients corresponding to each of the N target images, the image to be encoded corresponding to each target image
  • the first determining unit is specifically configured to determine a sum of each transform coefficient in the second set of transform coefficients The difference value for the image.
  • the first determining unit is specifically configured to determine N target values, where the second group of transform coefficients Divided into N subgroups, an ith target value of the N target values is a sum of transform coefficients in the ith subgroup; and the N target values are determined according to weighting coefficients of the N target values get on Weighted summation to obtain the image difference value.
  • the first determining unit is specifically configured to determine a first target value, where the first target value is a sum of significant transform coefficients of the second set of transform coefficients; determining a second target value, wherein the second target value is a sum of residual transform coefficients of the transform coefficient set of each image other than the second set of transform coefficients And weighting the first target value and the second target value according to the weighting coefficients of the first target value and the second target value to obtain the image difference value.
  • the second determining unit is specifically configured to use the image to be encoded and the image of the multiple candidate images a difference value, one candidate image having the smallest image difference value is selected from the plurality of candidate images; and a pixel value of the candidate image having the smallest difference value is determined as a pixel predicted value of the image to be encoded.
  • the second determining unit is specifically configured to use the image to be encoded and the image of the multiple candidate images a difference value, the E candidate images having the smallest image difference value are selected from the plurality of candidate images, E ⁇ 2; according to the image difference value of the image to be encoded and each of the E candidate images, Determining a weight of each candidate image; performing weighted averaging on pixel values of the E candidate image according to weights of the E candidate images to obtain a pixel prediction value of the image to be encoded.
  • the encoder further includes: a writing unit, configured to use the N candidate images in the knowledge The index in the library is written to the stream.
  • a decoder for selecting N candidate images from a knowledge base, wherein the N candidate images are used to determine pixel prediction values of an image to be decoded, and each candidate image and the candidate image
  • the shape and the size of the decoded image are the same
  • the image to be decoded includes at least one image block to be decoded, and the same transformation is performed on each of the image to be decoded and each of the N target images to obtain a transformation of the image to be decoded.
  • the N The target image is the N candidate images, or the N target images are images obtained by respectively obtaining the difference between the N candidate images and the image to be decoded at a corresponding pixel, and the transform coefficient set of the image to be decoded Corresponding to the transform coefficients in the transform coefficient set of each target image; selecting a transform coefficient whose amplitude satisfies a preset threshold from the transform coefficient set of the image to be decoded, to obtain a first set of transform coefficients; a set of transform coefficients, selecting a second set of transform coefficients corresponding to the first set of transform coefficients from the transform coefficient set of each target image; determining according to the second set of transform coefficients corresponding to the N target images Determining a pixel difference value between the image to be decoded and the N candidate images; determining a pixel prediction value of the image to be decoded according to the image difference value of the image to be decoded and the N candidate images
  • the N target images are the N candidate images
  • the encoder is specifically configured to use the first set of transform coefficients and each candidate image. And corresponding to the second set of transform coefficients, determining image difference values of the image to be decoded and each of the candidate images.
  • the encoder is specifically configured to determine the first group of transform coefficients and the second group of transform coefficients And a difference value of each corresponding transform coefficient; summing the difference values of the corresponding transform coefficients to obtain the image difference value.
  • the encoder is specifically configured to determine N target values, where the first group of transform coefficients and The second set of transform coefficients are divided into N subgroups in the same manner, and an ith target value among the N target values is an ith subgroup and a second set of transform coefficients in the first set of transform coefficients. And a sum of differences of the corresponding transform coefficients in the i-th sub-group; and weighting the N target values according to respective weighting coefficients of the N target values to obtain the image difference value.
  • the encoder is specifically configured to determine a first target value, where the first target value is the The sum of the difference between the first set of transform coefficients and each of the corresponding transform coefficients of the second set of transform coefficients; a second target value, wherein the second target value is a set of transform coefficients of the image to be decoded, except for the first set of transform coefficients, and a transform coefficient set of each candidate image a sum of differences of the respective transform coefficients of the remaining transform coefficients outside the second set of transform coefficients; and the first target value and the first according to the weighting coefficients of the first target value and the second target value The two target values are weighted and summed to obtain the image difference value.
  • the N target images are the N candidate images and the to-be-decoded image respectively at corresponding pixel points
  • the encoder is specifically configured to determine, according to the second set of transform coefficients corresponding to each target image of the N target images, the candidate image corresponding to the image to be decoded and each target image And an image difference value, wherein the candidate image corresponding to each target image and the image to be decoded are obtained at a corresponding pixel point to obtain each of the target images.
  • the encoder is specifically configured to determine a sum of each of the transform coefficients in the second set of transform coefficients as The image difference value.
  • the encoder is specifically configured to determine N target values, where the second set of transform coefficients are divided N subgroups, an ith target value of the N target values is a sum of transform coefficients in the ith subgroup; weighting the N target values according to weighting coefficients of the N target values Summing, the image difference value is obtained.
  • the encoder is specifically configured to determine a first target value, and the first target value is the second a sum of significant transform coefficients in the set of transform coefficients; determining a second target value, wherein the second target value is a sum of the remaining transform coefficients of the transform coefficient set of each image except the second set of transform coefficients; a weighting coefficient of the first target value and the second target value, and weighting the first target value and the second target value to obtain the image difference value.
  • the encoder is specifically configured to select one candidate image having the smallest image difference value from the plurality of candidate images according to the image difference value of the image to be decoded and the plurality of candidate images; The pixel value of the candidate image having the smallest difference value is determined as the pixel predicted value of the image to be decoded.
  • the encoder is specifically configured to: according to the image difference value between the image to be decoded and the plurality of candidate images Deselecting E candidate images having the smallest image difference value from the plurality of candidate images, E ⁇ 2; determining the image according to the image difference value of the image to be decoded and each of the E candidate images Calculating a weight of each candidate image; performing weighted averaging on pixel values of the E candidate image according to weights of the E candidate images to obtain a pixel prediction value of the image to be decoded.
  • the encoder is specifically configured to acquire the N candidate images from the code stream in the knowledge base.
  • An index of the N candidate images selected from the knowledge base according to the index.
  • an encoder for selecting N candidate images from a knowledge base, wherein the N candidate images are used to determine pixel prediction values of an image to be encoded, and each candidate image and the candidate image
  • the shape and size of the encoded image are the same
  • the image to be encoded includes at least one image block to be encoded; and the same transformation is performed on each of the image to be encoded and each of the N target images to obtain a transformation of the image to be encoded.
  • the transform coefficient set of the image to be encoded is in one-to-one correspondence with the transform coefficient of the transform coefficient set of each target image; and the amplitude of the transform coefficient set from the image to be encoded is selected to satisfy Presetting a transform coefficient of the threshold to obtain a first set of transform coefficients; selecting, according to the first set of transform coefficients, a set of transform coefficients from each of the target images and the first set a second set of transform coefficients corresponding to the transform coefficients; determining, according to the second set of transform coefficients corresponding to the N target images, image difference values of the image to be encoded and the N candidate images; And determining, by the image difference value of the N candidate images, a pixel prediction value of the image to be encoded; and encoding
  • the N target images are the N candidate images
  • the encoder is specifically configured to use the first set of transform coefficients and each candidate image.
  • the encoder is specifically configured to determine the first set of transform coefficients and the second set of transform coefficients And a difference value of each corresponding transform coefficient; summing the difference values of the corresponding transform coefficients to obtain the image difference value.
  • the encoder is specifically configured to determine N target values, where the first group of transform coefficients and The second set of transform coefficients are divided into N subgroups in the same manner, and an ith target value among the N target values is an ith subgroup and a second set of transform coefficients in the first set of transform coefficients. And a sum of differences of the corresponding transform coefficients in the i-th sub-group; and weighting the N target values according to respective weighting coefficients of the N target values to obtain the image difference value.
  • the encoder is specifically configured to determine a first target value, where the first target value is a sum of a difference between a first set of transform coefficients and respective ones of the second set of transform coefficients; determining a second target value, wherein the second target value is a transform coefficient set of the image to be encoded a sum of a residual transform coefficient outside the first set of transform coefficients and a difference between the transform coefficients of each of the candidate images and a corresponding transform coefficient of the remaining transform coefficients other than the second set of transform coefficients; And a weighting coefficient of the target value and the second target value, and weighting the first target value and the second target value to obtain the image difference value.
  • the N target images are the N candidate images and the to-be-coded image respectively at corresponding pixel points
  • the encoder is specifically configured to determine the image to be encoded and each target image according to a second set of transform coefficients corresponding to each target image in the N target images
  • the encoder is specifically configured to determine a sum of each of the transform coefficients in the second set of transform coefficients as The image difference value.
  • the encoder is specifically configured to determine N target values, where the second set of transform coefficients are divided N subgroups, an ith target value of the N target values is a sum of transform coefficients in the ith subgroup; weighting the N target values according to weighting coefficients of the N target values Summing, the image difference value is obtained.
  • the encoder is specifically configured to determine a first target value, where the first target value is the second a sum of significant transform coefficients in the set of transform coefficients; determining a second target value, wherein the second target value is a sum of the remaining transform coefficients of the transform coefficient set of each image except the second set of transform coefficients; a weighting coefficient of the first target value and the second target value, and weighting the first target value and the second target value to obtain the image difference value.
  • the encoder is specifically configured to use an image difference value between the image to be encoded and the plurality of candidate images. And selecting, from the plurality of candidate images, one candidate image having the smallest image difference value; determining a pixel value of the candidate image having the smallest difference value as a pixel predicted value of the image to be encoded.
  • the encoder is specifically configured to use an image difference value between the image to be encoded and the plurality of candidate images. And selecting, from the plurality of candidate images, E candidate images having the smallest image difference value, E ⁇ 2; determining, according to the image difference value of the image to be encoded and each of the E candidate images Calculating a weight of each candidate image; performing weighted averaging on pixel values of the E candidate image according to weights of the E candidate images to obtain a pixel prediction value of the image to be encoded.
  • the encoder is further configured to write an index of the N candidate images in the knowledge base Code stream.
  • the transform coefficient whose amplitude satisfies the preset threshold condition can better reflect the main structural information of the image, and the image difference value determined by the transform coefficient whose amplitude satisfies the preset threshold condition can reflect the image well.
  • the degree of difference between the two makes the subsequent image processing based on the image difference value more accurate.
  • FIG. 1 is a schematic flowchart of a method for processing an image according to an embodiment of the present invention.
  • FIG. 2 is an exemplary diagram of a target template image positional relationship of a prediction block.
  • FIG. 3 is a schematic block diagram of an apparatus for processing an image according to an embodiment of the present invention.
  • FIG. 4 is a schematic block diagram of an apparatus for processing an image according to an embodiment of the present invention.
  • FIG. 1 is a schematic flowchart of an image processing method according to an embodiment of the present invention.
  • the method of Figure 1 includes:
  • the difference image of the first image and the second image may specifically refer to an image obtained by the difference between the first image and the second image at the corresponding pixel point.
  • first image and the second image may be image regions of the same shape, the same size, including the same number of pixels, the shape of which may be, for example, a rectangle, an L shape, a triangle, a diamond, a trapezoid, a hexagon, or the like.
  • the first image may be a predicted image of a decoding prediction unit, or a reconstructed image of a template region corresponding to the decoding prediction unit, or a reconstructed image of a preset shape in the decoded image.
  • the second image may be a high quality image in a decoding end knowledge base, such as an original image pre-stored in the knowledge base; the second image may also be a local area in a decoded image.
  • the first image and the second image are respectively subjected to the same transform to obtain a first transform coefficient set and a second transform coefficient set.
  • the transformation here can adopt a variety of commonly used transformation methods, such as Discrete Cosine Transform (DCT), Discrete Sine Transform (DST), Hadamard Transform, Wavelet Transform ( Wavelet Transform), Scale-invariant Feature Transform (SIFT) transform, and the like.
  • DCT transform when the image is a rectangular image, the image can be transformed into a two-dimensional transformation matrix by two-dimensional DCT transformation, or the pixels in the image can be arranged into a one-dimensional vector, and the image is transformed into one-dimensional by one-dimensional DCT transformation.
  • the image is a triangle image
  • the pixels in the image may be arranged into a one-dimensional vector, one-dimensional DCT transformation may be performed, or the image may be rearranged into a rectangle to perform two-dimensional DCT transformation.
  • the transform coefficient set is a set of transform coefficients obtained by transforming the image, and may include all transform coefficients obtained by transforming the image, or may include some coefficients of all transform coefficients, for example, transform coefficients of a plurality of preset coefficient positions.
  • the first transform coefficient set and the second transform coefficient set described above may use the same transform coefficient numbering manner.
  • the numbering can be in a variety of forms, such as when transforming using a two-dimensional DCT transform
  • the transform coefficients may be numbered in the ZigZag scan order, or the transform coefficients may be numbered in order from left to right and top to bottom.
  • one or more color space components of the pixel may be transformed.
  • the color space components for example, the luminance component
  • an image difference value may be obtained; or a plurality of color space components may be separately transformed, and then multiple image difference values may be obtained, and then The image difference values under each color space component are averaged to obtain a final image difference value; of course, all the color space components can be combined and uniformly transformed, and then an image difference value is obtained.
  • the preset threshold condition may be that the amplitude of each transform coefficient in the first set of transform coefficients is not less than the amplitude of the remaining transform coefficients in the first transform coefficient set except the first set of transform coefficients or the first The amplitude of any one of a set of transform coefficients is not less than a threshold corresponding to the transform coefficient;
  • the first set of transform coefficients may be obtained from the first transform coefficient set by using any one of the following two processing modes:
  • Processing mode 1 determining whether the amplitude of each transform coefficient C i in the first transform coefficient set exceeds a threshold TH i ; if the amplitude of the coefficient crosses the threshold, adding the coefficient to the first set of transform coefficients.
  • the threshold TH i for each transform coefficient comparison may be the same or different.
  • the threshold may be a preset constant, or may be x times the first image corresponding to the quantization step (x is a positive real number); when the first image is an image in the decoded image, the threshold may also correspond to the decoded image. Extracted from the code stream.
  • Processing method 2 Select the M coefficients with the largest amplitude in the first transform coefficient set as the significant transform coefficients.
  • M is a positive integer, for example M is 4, 8, 15, 16, 20 or 32.
  • the first group of transform coefficients is which of the first transform coefficient sets.
  • the number can be indicated by means of transform coefficient information.
  • the transform coefficient information may have various forms.
  • the transform coefficient information may be a number of the first set of transform coefficients in the first transform coefficient set; for example, the transform coefficient information may be an array, the dimension of the array and the first transform.
  • the number of transform coefficients in the coefficient set is the same, and each element in the array indicates whether one transform coefficient in the first transform coefficient set belongs to the first set of transform coefficients.
  • step 140 may include: converting the coefficients according to the first set of transform coefficients, and the first transform coefficient set and the second transform coefficient set a one-to-one correspondence, selecting a second set of transform coefficients from the second transform coefficient set; and determining an image difference value between the first image and the second image according to a difference between the first set of transform coefficients and the second set of transform coefficients.
  • the step 140 may include: according to the first set of transform coefficients, and the first transform coefficient set and the second transform a one-to-one correspondence of the transform coefficients of the coefficient set, and selecting a second set of transform coefficients from the second transform coefficient set; and determining a difference value between the first image and the second image according to the second set of transform coefficients.
  • the above processing may include at least one of the following processes:
  • the transform coefficient whose amplitude satisfies the preset threshold condition can better reflect the main structural information of the image, and the image difference value determined by the transform coefficient whose amplitude satisfies the preset threshold condition can reflect the image well.
  • the degree of difference between the two makes the subsequent image processing based on the image difference value more accurate.
  • the second transform coefficient set is a transform coefficient set obtained by transforming the second image
  • the step 140 may include: performing one-to-one correspondence between the first transform coefficient set and the second transform coefficient set transform coefficient. And selecting, from the second transform coefficient set, a second set of transform coefficients corresponding to the first set of transform coefficients; and determining an image difference value according to the first set of transform coefficients and the second set of transform coefficients.
  • determining, according to the first set of transform coefficients and the second set of transform coefficients, the image difference value may include: determining a difference between each of the first set of transform coefficients and the second set of transform coefficients The sum of the differences of the respective transform coefficients is determined as an image difference value.
  • weighting coefficients are real numbers, usually positive real numbers, and the weighting coefficients of the subgroups may be determined by the position of the transform coefficients in the subset in the transform coefficient set.
  • the transform of the first image is DCT or DST transform
  • all transform coefficients are numbered according to a common ZigZag scan order, and the number of values in the first transform coefficient set is smaller than the significant transform coefficient of the total number value 1/2 (ie, transform The transform coefficient whose amplitude satisfies the preset threshold condition is divided into subgroup 1 whose weighting coefficient is 0.75; the remaining significant transform coefficients in the first transform coefficient set divide subgroup 2, and the weighting coefficient is 0.25.
  • the significant transform coefficients in the 1/4 region of the upper left corner of the transform coefficient matrix in the first transform coefficient set are divided into subgroup 1 with a weighting coefficient of 1.2.
  • the weighting coefficient is 0.8; and the first transform coefficient is concentrated
  • the significant transform coefficients located in the 1/4 region of the lower right corner of the transform coefficient matrix are divided into subgroups 3 with a weighting coefficient of 0.25.
  • determining the image difference value according to the first set of transform coefficients and the second set of transform coefficients may include: determining a first target value, where the first target value is the first set of transform coefficients and the first a sum of difference values of respective transform coefficients of the two sets of transform coefficients; determining a second target value, wherein the second target value is a residual transform coefficient and a second transform coefficient set other than the first set of transform coefficients in the first transform coefficient set a sum of difference values of respective ones of the remaining transform coefficients except the second set of transform coefficients; weighting and summing the first target value and the second target value according to the weighting coefficients of the first target value and the second target value , get the image difference value.
  • the weighting coefficient of the first target value may be a positive real number, for example, 1. 0.5 or 2 may be selected, and the weighting coefficient of the second target value may be a non-zero real number, for example, -0.2, 0.5, or 0.8 may be selected;
  • the weighting coefficient of a target value and the weighting coefficient of the second target value may not be equal, and generally the weighting coefficient of the first target value is greater than the weighting coefficient of the second target value.
  • the sum of the differences mentioned above may be one of mean square error (MSE), mean absolute difference (MAD), squared error and (SSE), absolute error and (SAD), and may also be used.
  • Other summation calculation methods More specifically, for example, when the sum of the differences uses a squared error sum, the corresponding transform coefficients of the two sets of transform coefficients are subtracted by two to obtain a set of differences, and the squares of each difference are added to obtain a squared error sum; For example, when the sum of the differences uses the average absolute difference, the corresponding transform coefficients of the two sets of transform coefficients are subtracted to obtain a set of differences, and the absolute values of each difference are added to obtain an absolute error sum, and then the absolute error is obtained. Normalization is performed by dividing by the number of differences to obtain an average absolute difference.
  • the second transform coefficient set is a transform coefficient set obtained by transforming the difference image of the first image and the second image
  • the step 140 may include: according to the first transform coefficient set and the second transform. a one-to-one correspondence of the transform coefficients of the set of coefficients, selecting a second set of transform coefficients corresponding to the first set of transform coefficients from the second transform coefficient set; and determining an image difference value according to the second set of transform coefficients.
  • determining, according to the second set of transform coefficients, the image difference value includes: determining a sum of each transform coefficient in the second set of transform coefficients as an image difference value.
  • the second set of transform coefficients includes N subgroups
  • determining the image difference value according to the second set of transform coefficients may include: determining N target values, where the ith target value is the i th sub The sum of the transform coefficients in the group; weighting the N target values according to the weighting coefficients of the N target values to obtain an image difference value.
  • determining the image difference value according to the second set of transform coefficients may include: determining a first target value, where the first target value is a sum of transform coefficients of the second set of transform coefficients; determining second a target value, the second target value is a sum of the remaining transform coefficients except the second set of transform coefficients in the second transform coefficient set; and the first target value and the second target according to the weighting coefficients of the first target value and the second target value The values are weighted and summed to obtain image difference values.
  • the weighting coefficient of the first numerical sum may be a positive real number, for example, 1, 0.5 or 2, etc.
  • the weighting coefficient of the second numerical sum may be a non-zero real number, such as -0.2, 0.5, 0.8, or 0.9, etc.
  • the weighting coefficients of the sum of the weighting coefficients of the sum and the sum of the second values may not be equal, and generally the weighting coefficients of the sum of the first values are greater than the weighting coefficients of the sum of the second values.
  • the sum of the above transform coefficients may be a sum of absolute values, a sum of square values, and a flat One of the mean values of the square values, the mean of the absolute values, and the Root Mean Squared Error (RMSE), other forms of calculation can also be used. More specifically, for example, when the sum of the transform coefficients uses the sum of the absolute values, the absolute values of each coefficient are added to obtain the sum of the absolute values; and, for example, when the sum of the transform coefficients uses the mean of the squared values, the transform coefficients are The squared values are added to obtain the sum of the squared values, and the sum of the squared values is divided by the number of these transform coefficients to be normalized to obtain the mean of the squared values.
  • RMSE Root Mean Squared Error
  • the first image is a target image
  • the second image is any candidate image of the K candidate images
  • the step 140 may include: determining, according to the target image and the image difference value of each candidate image, A weighting coefficient of each candidate image is determined; a filtered image of the target image is determined according to pixel values of the K candidate images and weighting coefficients of the K candidate images.
  • the visual quality of the filtered image is higher than the target image, and the filtered image is replaced with the target image, which can enhance the visual quality of the image where the target image is located.
  • any one of the above K candidate images may have the same shape and size as the target image.
  • the target image may be an area in an image, such as an image of a preset shape in the decoded image, such as a rectangle or a triangle.
  • each of the target image and the K candidate images has one image difference value, and the manner of determining the image difference value between the target image and each candidate image may be the manner described in the above embodiment.
  • the pixel value may be one of the color space components of the pixel, and the pixel value is a scalar; the pixel value may also be a multi-dimensional vector composed of multiple components in the color space component of the pixel.
  • the color space of the pixel is a common color space such as RGB, YCbCr, YUV, HSV, CIE Lab.
  • the manner of determining the weighting coefficients of each candidate image according to the image difference value between the target image and each candidate image may be implemented in various manners. Several implementations are described below.
  • each candidate image is determined, wherein b1, a1, and h1 each represent a positive real number, Dk represents an image difference value of the target image and the kth candidate image, and Wk represents a weighting coefficient of the kth candidate image.
  • weighting coefficients of each candidate image are determined, wherein b3, a3, and h3 each represent a positive real number, Dk represents an image difference value of the target image and the kth candidate image, and Wk represents a weighting coefficient of the kth candidate image.
  • the manner of determining the filtered image of the target image according to the pixel values of the K candidate images and the weighting coefficients of the K candidate images may also be various, for example, according to Determining a filtered image of the target image, wherein Pf j represents the pixel value of the filtered image at the jth pixel, W0 represents the weighting coefficient of the target image, Wk represents the weighting coefficient of the kth candidate image, and P0 j represents the target image at the jth pixel The pixel value of the point, Pk j represents the pixel value of the kth candidate image at the jth pixel point.
  • the first image is a target image
  • the second image is any candidate image of the K candidate images
  • performing subsequent image processing on the first image and the second image according to the image difference value including: According to the image difference value of the target image and the K candidate images, G candidate images having the smallest image difference value from the target image are selected from the K candidate images as the nearest neighbor image of the target image, wherein G is less than or equal to K.
  • the target image may be an image region in an image, such as an image of a preset shape in the decoded image, and the target image may also be a predicted image of a decoding prediction unit, for example, by motion vector based inter prediction in H.265.
  • the obtained predicted image; the preset shape is, for example, a rectangle or a triangle.
  • the K candidate images may be images in the reconstructed region of the currently decoded image or images in other decoded images, or may be images in the decoding end knowledge base.
  • each of the target image and the K candidate images has an image difference value
  • the manner in which the image difference value of the target image and each candidate image is determined may adopt the image difference value determination manner in any of the above embodiments.
  • the nearest neighbor image may be used to enhance the target image, for example, the nearest neighbor image with the smallest image difference value is replaced with the target image; and the target image is filtered by, for example, the nearest neighbor image, and the G nearest neighbor images and targets are used. The images are combined together to form a filtered image to replace the target image P0.
  • the nearest neighbor image can also be used to indicate the starting position of the iterative search. For example, first find the nearest neighbor image in several images around the starting search point, and then use the nearest neighbor image as the new search starting point for the next iterative search. .
  • the first image is a target template image corresponding to the prediction block, where the prediction block corresponds to K candidate blocks, the K candidate blocks respectively correspond to K candidate template images, and the second image is K candidates.
  • the template image is respectively subjected to subsequent residual image processing of the first image and the second image according to the image difference value, and any residual template image in the K residual template images obtained by the difference between the template image and the target template image, including: The image difference value between the target template image and the K residual template images, and the G residual template images having the smallest image difference value from the target template image are selected from the K residual template images, wherein G is less than or equal to K;
  • G candidate blocks corresponding to the G residual template images are selected; and pixel prediction values of the prediction block are determined according to pixel values of the G candidate blocks.
  • the first image may be a target template image corresponding to the prediction block
  • the third image may be a K residual template image in which the K candidate template images are respectively obtained by subtracting the target template image from the corresponding pixel points. Any residual template image.
  • the prediction block is usually a rectangular area, which may have the same shape and size as any one of the candidate blocks; the target template image and any one candidate template image may have the same shape and size; the prediction block and the target template image form an area and any pair
  • the regions formed by the candidate blocks and the candidate template images may have the same shape and size.
  • the target template image may be adjacent to the prediction block, usually a rectangular area or an L-shaped area, or may be other shapes.
  • FIG. 2 shows several examples of the target template image T0 of the prediction block B0.
  • the candidate template image may be an image in the decoding end knowledge base, or may be an image in the reconstructed area of the currently decoded image or an image in other decoded images.
  • Determining the pixel prediction value of the prediction block according to the pixel values of the G candidate blocks may include: Determining a pixel prediction value of the prediction block, where B0 j represents a pixel value of the jth pixel of the prediction block, Bg j represents a pixel value of the jth candidate block at the jth pixel point, and Wg represents a weighting coefficient corresponding to the gth candidate block .
  • the weighting coefficient Wg corresponding to the candidate block may be determined according to the image difference value Dg of the candidate template image of the candidate block and the target template image. For example, the weighting coefficient Wg decreases as the Dg increases, specifically, according to The weighting coefficients corresponding to the G candidate blocks are determined, wherein b1, a1, and h1 each represent a positive real number, Dg represents an image difference value between the target image and the gth candidate image, and Wg represents a weighting coefficient corresponding to the gth candidate block.
  • the image difference value, Wg represents the weighting coefficient of the gth candidate image.
  • FIGS. 3 and 4 can implement the various steps of the method of processing an image described in FIG. 1, and the repeated description is appropriately omitted for the sake of brevity.
  • FIG. 3 is a schematic block diagram of an apparatus for processing an image according to an embodiment of the present invention.
  • the apparatus 300 of Figure 3 includes:
  • a first transform unit 310 configured to transform the first image to obtain a first transform coefficient set
  • a second transform unit 320 configured to perform the transform on the second image, or perform the transform on the difference image of the first image and the second image to obtain a second transform coefficient set
  • the selecting unit 330 is configured to select, according to the amplitude of the transform coefficient in the first transform coefficient set obtained by the first transform unit 310, a first set of transform coefficients from the first transform coefficient set, where the first The amplitude of the transform coefficients in the set of transform coefficients satisfies a preset threshold condition;
  • a determining unit 340 configured to determine, according to the first set of transform coefficients selected by the selecting unit 330 and the second set of transform coefficients obtained by the second transforming unit 320, between the first image and the second image Image difference value;
  • the processing unit 350 is configured to process the first image and the second image according to the image difference value determined by the determining unit 340.
  • the transform coefficient whose amplitude satisfies the preset threshold condition can better reflect the main structural information of the image, and the image difference value determined by the transform coefficient whose amplitude satisfies the preset threshold condition can reflect the image well.
  • the degree of difference between the two makes the subsequent image processing based on the image difference value more accurate.
  • the second set of transform coefficients is a transform coefficient set obtained by performing the transform on the second image, where the determining unit 340 is specifically configured to use, according to the first transform coefficient set
  • the second transform coefficient sets a one-to-one correspondence of the transform coefficients, and selects, according to the second transform coefficient set, a second set of transform coefficients corresponding to the first set of transform coefficients; according to the first set of transform coefficients and The second set of transform coefficients is described to determine the image difference value.
  • the determining unit 340 is specifically configured to determine a difference between each of the first set of transform coefficients and each of the second set of transform coefficients; The sum of the differences of the coefficient of change is determined as the image difference value.
  • the first group of transform coefficients and the second set of transform coefficients are all divided into N subgroups in the same manner, and the determining unit 340 is specifically configured to determine N target values, where An ith target value of the N target values is a sum of difference values of respective transform coefficients of an i th subgroup of the first set of transform coefficients and an i th subgroup of the second set of transform coefficients And weighting the N target values according to the weighting coefficients of the N subgroups to obtain the image difference value.
  • the determining unit 340 is specifically configured to determine a first target value, where the first target value is corresponding to each of the first set of transform coefficients and the second set of transform coefficients. a sum of difference values of the transform coefficients; the second target value is determined, wherein the second target value is a residual transform coefficient and the second transform coefficient except the first set of transform coefficients in the first transform coefficient set And a sum of difference values of respective transform coefficients of the remaining transform coefficients except the second set of transform coefficients; and the first target value according to the first target value and the weighting coefficient of the second target value Performing weighted summation with the second target value to obtain the image difference value.
  • the second set of transform coefficients is a transform coefficient set obtained by performing the transform on the difference image of the first image and the second image, where the determining unit 340 is specifically used. And selecting, according to the one-to-one correspondence between the first transform coefficient set and the second transform coefficient set transform coefficient, selecting a second set of transform coefficients corresponding to the first set of transform coefficients from the second transform coefficient set; The image difference value is determined based on the second set of transform coefficients.
  • the determining unit 340 is specifically configured to determine, as the image difference value, a sum of transform coefficients in the second set of transform coefficients.
  • the second group of transform coefficients includes N subgroups
  • the determining unit 340 is specifically configured to determine N target values, where the i th target value is a transform coefficient in the i th subgroup. And summing the N target values according to the weighting coefficients of the N target values to obtain the image difference value.
  • the determining unit 340 is specifically configured to determine a first target value,
  • the first target value is a sum of transform coefficients of the second set of transform coefficients; determining a second target value, where the second target value is in addition to the second set of transform coefficients And a sum of the remaining transform coefficients; and weighting the first target value and the second target value according to the weighting coefficients of the first target value and the second target value to obtain the image difference value .
  • the first image is a target image
  • the second image is any candidate image among the K candidate images
  • the processing unit 350 is specifically configured to use the target image and each An image difference value of the candidate image, determining a weighting coefficient of each of the candidate images; determining a filtered image of the target image according to pixel values of the K candidate images and weighting coefficients of the K candidate images.
  • the determining unit 340 is specifically configured to Determining weighting coefficients of each of the candidate images, wherein b1, a1, and h1 each represent a positive real number, Dk represents an image difference value of the first image and the kth candidate image, and Wk represents the kth candidate image Weighting factor.
  • Dk represents an image difference value of the first image and the kth candidate image
  • Wk represents a weighting coefficient of the kth candidate image.
  • the determining unit 340 is specifically configured to Determining weighting coefficients of each of the candidate images, wherein b3, a3, and h3 each represent a positive real number, Dk represents an image difference value of the first image and the kth candidate image, and Wk represents the kth candidate image Weighting factor.
  • the processing unit 350 is specifically configured to Determining the filtered image, wherein Pf j represents a pixel value of the filtered image at a j-th pixel point, W0 represents a weighting coefficient of the first image, Wk represents a weighting coefficient of the k-th candidate image, and P0 j represents The pixel value of the first image at the jth pixel point, and Pk j represents the pixel value of the kth candidate image at the jth pixel point.
  • FIG. 4 is a schematic block diagram of an apparatus for processing an image according to an embodiment of the present invention.
  • the apparatus 400 of Figure 4 includes:
  • a memory 410 configured to store a program
  • a processor 420 configured to execute a program, when the program is executed, the processor 420 is specifically configured to: transform a first image to obtain a first transform coefficient set; perform the transform on the second image, or Performing the transform on the difference image of the first image and the second image to obtain a second transform coefficient set; selecting the first group from the first transform coefficient set according to the magnitude of the transform coefficient in the first transform coefficient set a transform coefficient, wherein an amplitude of the transform coefficients in the first set of transform coefficients satisfies a preset threshold condition; determining the first image and the second image according to the first set of transform coefficients and the second set of transform coefficients An image difference value between; the first image and the second image are processed according to the image difference value.
  • the transform coefficient whose amplitude satisfies the preset threshold condition can better reflect the main structural information of the image, and the image difference value determined by the transform coefficient whose amplitude satisfies the preset threshold condition can reflect the image well.
  • the degree of difference between the two makes the subsequent image processing based on the image difference value more accurate.
  • the second set of transform coefficients is a transform coefficient set obtained by performing the transform on the second image
  • the processor 420 is specifically configured to use the first transform coefficient set according to The second transform coefficient sets a one-to-one correspondence of the transform coefficients, and selects, according to the second transform coefficient set, a second set of transform coefficients corresponding to the first set of transform coefficients; according to the first set of transform coefficients and The second set of transform coefficients is described to determine the image difference value.
  • the processor 420 is specifically configured to determine a difference between each of the first set of transform coefficients and the second set of transform coefficients, where the corresponding transform coefficients are The sum of the differences is determined as the image difference value.
  • the first set of transform coefficients and the second set of transform coefficients Each of the N target values is the ith of the first set of transform coefficients. And a sum of differences between the subgroups and the corresponding transform coefficients in the i th subgroup of the second set of transform coefficients; and weighting and summing the N target values according to the weighting coefficients of the N subgroups The image difference value.
  • the processor 420 is specifically configured to determine a first target value, where the first target value is corresponding to each of the first set of transform coefficients and the second set of transform coefficients a sum of difference values of the transform coefficients; the second target value is determined, wherein the second target value is a residual transform coefficient and the second transform coefficient except the first set of transform coefficients in the first transform coefficient set And a sum of difference values of respective transform coefficients of the remaining transform coefficients except the second set of transform coefficients; and the first target value according to the first target value and the weighting coefficient of the second target value Performing weighted summation with the second target value to obtain the image difference value.
  • the second transform coefficient set is a transform coefficient set obtained by performing the transform on the difference image of the first image and the second image, where the processor 420 is specifically used. And selecting, according to the one-to-one correspondence between the first transform coefficient set and the second transform coefficient set transform coefficient, selecting a second set of transform coefficients corresponding to the first set of transform coefficients from the second transform coefficient set; The image difference value is determined based on the second set of transform coefficients.
  • the processor 420 is specifically configured to determine, as the image difference value, a sum of transform coefficients of the second set of transform coefficients.
  • the second group of transform coefficients includes N subgroups
  • the processor 420 is specifically configured to determine N target values, where the i th target value is a transform coefficient in the i th subgroup. And summing the N target values according to the weighting coefficients of the N target values to obtain the image difference value.
  • the processor 420 is specifically configured to determine a first target value, where the first target value is a sum of transform coefficients of the second set of transform coefficients; and determining a second target value, The second target value is a remainder of the second transform coefficient set except the second set of transform coefficients And a sum of the transform coefficients; and weighting the first target value and the second target value according to the weighting coefficients of the first target value and the second target value to obtain the image difference value.
  • the first image is a target image
  • the second image is any candidate image among the K candidate images
  • the processor 420 is specifically configured to use the target image and each An image difference value of the candidate image, determining a weighting coefficient of each of the candidate images; determining a filtered image of the target image according to pixel values of the K candidate images and weighting coefficients of the K candidate images.
  • the processor 420 is specifically configured to Determining weighting coefficients of each of the candidate images, wherein b1, a1, and h1 each represent a positive real number, Dk represents an image difference value of the first image and the kth candidate image, and Wk represents the kth candidate image Weighting factor.
  • Dk represents an image difference value of the first image and the kth candidate image
  • Wk represents a weighting coefficient of the kth candidate image.
  • the processor 420 is specifically configured to Determining weighting coefficients of each of the candidate images, wherein b3, a3, and h3 each represent a positive real number, Dk represents an image difference value of the first image and the kth candidate image, and Wk represents the kth candidate image Weighting factor.
  • the processor 420 is specifically configured to Determining the filtered image, wherein Pf j represents a pixel value of the filtered image at a jth pixel point, W0 represents a weighting coefficient of the first image, Wk represents a weighting coefficient of the kth candidate image, and P0 j represents The pixel value of the first image at the jth pixel point, and Pk j represents the pixel value of the kth candidate image at the jth pixel point.
  • the term "and/or” is merely an association relationship describing an associated object, indicating that there may be three relationships.
  • a and/or B may indicate that A exists separately, and A and B exist simultaneously, and B cases exist alone.
  • the character "/" in this article generally indicates that the contextual object is an "or" relationship.
  • the disclosed systems, devices, and methods may be implemented in other manners.
  • the device embodiments described above are merely illustrative.
  • the division of the unit is only a logical function division.
  • there may be another division manner for example, multiple units or components may be combined or Can be integrated into another system, or some features can be ignored or not executed.
  • the mutual coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interface, device or unit, or an electrical, mechanical or other form of connection.
  • the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, may be located in one place, or may be distributed to multiple network units. Some or all of the units may be selected according to actual needs to achieve the objectives of the embodiments of the present invention.
  • each functional unit in various embodiments of the present invention may be integrated in one processing unit. It is also possible that each unit physically exists alone, or two or more units may be integrated in one unit.
  • the above integrated unit can be implemented in the form of hardware or in the form of a software functional unit.
  • the integrated unit if implemented in the form of a software functional unit and sold or used as a standalone product, may be stored in a computer readable storage medium.
  • the technical solution of the present invention contributes in essence or to the prior art, or all or part of the technical solution may be embodied in the form of a software product stored in a storage medium.
  • a number of instructions are included to cause a computer device (which may be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods described in various embodiments of the present invention.
  • the foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, and the like. .

Landscapes

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

Abstract

一种处理图像的方法和装置，该方法包括：对第一图像进行变换，得到第一变换系数集；对第二图像进行变换，或者对第一图像和第二图像的差值图像进行变换，得到第二变换系数集；根据第一变换系数集中的变换系数的幅度，从第一变换系数集中选取第一组变换系数，其中，第一组变换系数中的变换系数的幅度满足预设的阈值条件；根据第一组变换系数和第二变换系数集，确定第一图像和第二图像之间的图像差异值；根据图像差异值，对第一图像和第二图像进行处理。实施例确定的图像差异值能够较好地反映两个图像的差异程度，使得基于该图像差异值的后续图像处理更加准确。

Description

处理图像的方法和装置
本申请要求于2015年01月09日提交中国专利局、申请号为201510012142.4、发明名称为“处理图像的方法和装置”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本发明涉及领域图像处理领域，并且更具体地，涉及处理图像的方法和装置。
背景技术
视频编码，例如H.264/AVC、H.265/HEVC、AVS等视频编码标准，通常采用混合编码框架，主要包括预测(prediction)、变换(transform)、量化(quantization)、熵编码(entropy coding)等环节。视频解码是将码流转换为视频图像的过程，它包括熵解码(entropy decoding)、预测、反量化(dequantization)、反变换(inverse transform)等几个主要环节。首先，将码流通过熵解码处理解析出编码模式信息和量化后的变换系数。然后，一方面由编码模式信息和已经解码的重建像素得出预测像素；另一方面将量化后的变换系数通过反量化得到重建的变换系数，再对重建的变换系数进行反变换，得到重建的残差信息。之后，将重建的残差信息和预测像素相加，得到重建像素，从而恢复出视频图像。
对于有损编码，重建像素(reconstructed pixel)与原始像素(original pixel)可能是不相同的，两者之间的数值差异称为失真(distortion)。一般来说，失真由量化引起，量化参数(Quantization parameter,QP)越大则失真越强，图像变得模糊，总得来说像素质量越差。
基于知识库的编码是H.264/AVC、H.265/HEVC的一种扩展。解码端包含一个知识库，其中存储了一些图像(image)或/和图像区域(image region)， 称之为图样(patch)。知识库中的图像或图样可以来自当前解码视频中已经解码的重建图像，例如从已解码的重建图像中提取一些代表性的图像加入知识库；知识库中的图像或图样也可以来自当前解码视频的重建图像之外，例如由解码其它视频得到的重建图像或图样，又例如解码系统预先存储的包含多个图像或图样，这些预先存储的图像或图样可以是没有经过编码压缩的原始图像。解码当前视频时，可以利用知识库中的像素信息。例如，解码时利用到的预测像素信息可以来自于知识库中的像素信息。
预测环节利用已编码区域的重建像素产生当前编码块对应的原始像素的预测像素(predicted pixel)。预测方式主要包括帧内预测(intra prediction)和帧间预测(inter prediction)两大类。其中，帧内编码中的模版匹配(template matching)和帧间编码中的解码端运动矢量导出(decoder side motion vector derivation)技术需要利用当前解码预测块周围的重建图像模版区域(template)，在当前帧中已重建区域或其它已重建帧中搜索与当前解码块模版区域差异最小的一个或多个最近邻(nearest neighbor)图像，称为匹配图像。对于这两类技术，怎样评价模版区域图像和匹配过程中的某个候选模版区域图像的图像差异值(difference)或者在值域空间的距离(distance)是一个关键问题，直接决定了最终的搜索结果。传统的两个图像之间的图像差异值计算方法例如两图像像素域的平方误差和(Sum of Squared Errors,SSE)、绝对误差和(Sum of Absolute Difference,SAD)、均方误差(Mean Squared Error,MSE)、平均绝对差(Mean Absolute Difference,MAD)，又例如两图像经过Hadamard变换后的变换系数域的平均绝对差(Sum of Absolute Transformed Difference，SATD)。图像差异值的计算问题也在其它图像搜索和图像融合处理中也起到了关键作用。传统的图像差异值计算方法会将高质量图像相对于低质量图像的信号质量提升错误地认为是两个图像之间的差异，利用两个图像像素域的平方误差和等传统差异计算方法得到的差异较小的图像很可能在视觉上并不相似，导致后续的图像处理结果不准确。
发明内容
本发明实施例提供了一种处理图像的方法和装置，以提高图像处理的准确性。
第一方面，提供一种处理图像的方法，包括：对第一图像进行变换，得到第一变换系数集；对第二图像进行所述变换，或者对第一图像和第二图像的差值图像进行所述变换，得到第二变换系数集；根据所述第一变换系数集中的变换系数的幅度，从所述第一变换系数集中选取第一组变换系数，其中，所述第一组变换系数中的变换系数的幅度满足预设的阈值条件；根据所述第一组变换系数和所述第二变换系数集，确定第一图像和第二图像之间的图像差异值；根据所述图像差异值，对所述第一图像和所述第二图像进行处理。
结合第一方面，在第一方面的一种实现方式中，所述第二变换系数集是对所述第二图像进行所述变换得到的变换系数集，所述根据所述第一组变换系数和所述第二变换系数集，确定第一图像和第二图像之间的图像差异值，包括：根据所述第一变换系数集与所述第二变换系数集中变换系数的一一对应关系，从所述第二变换系数集中选取与所述第一组变换系数对应的第二组变换系数；根据所述第一组变换系数和所述第二组变换系数，确定所述图像差异值。
结合第一方面或其上述实现方式的任一种，在第一方面的另一种实现方式中，所述根据所述第一组变换系数和所述第二组变换系数，确定所述图像差异值，包括：确定所述第一组变换系数和所述第二组变换系数中各对应变换系数的差值；将所述各对应变换系数的差值之和确定为所述图像差异值。
结合第一方面或其上述实现方式的任一种，在第一方面的另一种实现方式中，所述第一组变换系数和所述第二组变换系数均按照相同方式划分为N个子组，所述根据所述第一组变换系数和所述第二组变换系数，确定所述图像差异值，包括：确定N个目标值，其中，所述N个目标值中的第i目标值 为所述第一组变换系数中的第i子组与所述第二组变换系数中的第i子组中各对应变换系数的差值之和；根据所述N个子组的加权系数，对所述N个目标值进行加权求和，得到所述图像差异值。
结合第一方面或其上述实现方式的任一种，在第一方面的另一种实现方式中，所述根据所述第一组变换系数和所述第二组变换系数，确定所述图像差异值，包括：确定第一目标值，其中，所述第一目标值为所述第一组变换系数与所述第二组变换系数中各对应变换系数的差值之和；确定第二目标值，其中，所述第二目标值为所述第一变换系数集中除所述第一组变换系数外的剩余变换系数与所述第二变换系数集中除所述第二组变换系数外的剩余变换系数中各对应变换系数的差值之和；根据所述第一目标值和所述第二目标值的加权系数，对所述第一目标值和所述第二目标值进行加权求和，得到所述图像差异值。
结合第一方面或其上述实现方式的任一种，在第一方面的另一种实现方式中，所述第二变换系数集是对所述第一图像和所述第二图像的差值图像进行所述变换得到的变换系数集，所述根据所述第一组变换系数和所述第二变换系数集，确定第一图像和第二图像之间的图像差异值，包括：根据所述第一变换系数集与所述第二变换系数集中变换系数的一一对应关系，从所述第二变换系数集中选取与所述第一组变换系数对应的第二组变换系数；根据所述第二组变换系数，确定所述图像差异值。
结合第一方面或其上述实现方式的任一种，在第一方面的另一种实现方式中，所述根据所述第二组变换系数，确定所述图像差异值，包括：将所述第二组变换系数中各变换系数之和确定为所述图像差异值。
结合第一方面或其上述实现方式的任一种，在第一方面的另一种实现方式中，所述第二组变换系数包括N个子组，所述根据所述第二组变换系数，确定所述图像差异值，包括：确定N个目标值，其中，第i目标值为第i子组中的变换系数之和；根据所述N个目标值的加权系数，对所述N个目标 值进行加权求和，得到所述图像差异值。
结合第一方面或其上述实现方式的任一种，在第一方面的另一种实现方式中，所述根据所述第二组变换系数，确定所述图像差异值，包括：确定第一目标值，所述第一目标值为所述第二组变换系数中各变换系数之和；确定第二目标值，所述第二目标值为所述第二变换系数集中除所述第二组变换系数外的剩余变换系数之和；根据所述第一目标值和所述第二目标值的加权系数，对所述第一目标值和所述第二目标值进行加权求和，得到所述图像差异值。
结合第一方面或其上述实现方式的任一种，在第一方面的另一种实现方式中，所述第一图像为目标图像，所述第二图像为K个候选图像中的任意候选图像，所述根据所述图像差异值，对所述第一图像和所述第二图像进行处理，包括：根据所述目标图像和每个候选图像的图像差异值，确定所述每个候选图像的加权系数；根据所述K个候选图像的像素值，以及所述K个候选图像的加权系数，确定所述目标图像的滤波图像。
结合第一方面或其上述实现方式的任一种，在第一方面的另一种实现方式中，所述根据所述目标图像和每个候选图像的图像差异值，确定所述每个候选图像的加权系数，包括：根据
Figure PCTCN2016070228-appb-000001
确定所述每个候选图像的加权系数，其中，b1、a1和h1均表示正实数，Dk表示所述第一图像和第k个候选图像的图像差异值，Wk表示所述第k个候选图像的加权系数。
结合第一方面或其上述实现方式的任一种，在第一方面的另一种实现方式中，所述根据所述目标图像和每个候选图像的图像差异值，确定所述每个候选图像的加权系数，包括：根据Wk＝b2-(Dk)a2/h2确定所述每个候选图像的加权系数，其中，b2、a2和h2均表示正实数，Dk表示所述第一图像和第k个候选图像的图像差异值，Wk表示所述第k个候选图像的加权系数。
结合第一方面或其上述实现方式的任一种，在第一方面的另一种实现方式中，所述根据所述目标图像和每个候选图像的图像差异值，确定所述每个 候选图像对应的加权系数，包括：根据
Figure PCTCN2016070228-appb-000002
确定所述每个候选图像的加权系数，其中，b3、a3和h3均表示正实数，Dk表示所述第一图像和第k个候选图像的图像差异值，Wk表示所述第k个候选图像的加权系数。
结合第一方面或其上述实现方式的任一种，在第一方面的另一种实现方式中，所述根据所述K个候选图像的像素值，以及所述K个候选图像的加权系数，确定所述目标图像的滤波图像，包括：根据
Figure PCTCN2016070228-appb-000003
确定所述滤波图像，其中，Pfj表示所述滤波图像在第j像素点的像素值，W0表示所述第一图像的加权系数，Wk表示第k个候选图像的加权系数，P0j表示所述第一图像在第j像素点的像素值，Pkj表示所述第k个候选图像在第j像素点的像素值。
第二方面，提供一种处理图像的装置，包括：第一变换单元，用于对第一图像进行变换，得到第一变换系数集；第二变换单元，用于对第二图像进行所述变换，或者对第一图像和第二图像的差值图像进行所述变换，得到第二变换系数集；选取单元，用于根据所述第一变换单元得到的所述第一变换系数集中的变换系数的幅度，从所述第一变换系数集中选取第一组变换系数，其中，所述第一组变换系数中的变换系数的幅度满足预设的阈值条件；确定单元，用于根据所述选取单元选取的所述第一组变换系数和所述第二变换单元得到的所述第二变换系数集，确定第一图像和第二图像之间的图像差异值；处理单元，用于根据所述确定单元确定的所述图像差异值，对所述第一图像和所述第二图像进行处理。
结合第二方面，在第二方面的一种实现方式中，所述第二变换系数集是对所述第二图像进行所述变换得到的变换系数集，所述确定单元具体用于根据所述第一变换系数集与所述第二变换系数集中变换系数的一一对应关系，从所述第二变换系数集中选取与所述第一组变换系数对应的第二组变换系 数；根据所述第一组变换系数和所述第二组变换系数，确定所述图像差异值。
结合第二方面或其上述实现方式的任一种，在第二方面的另一种实现方式中，所述确定单元具体用于确定所述第一组变换系数和所述第二组变换系数中各对应变换系数的差值；将所述各对应变换系数的差值之和确定为所述图像差异值。
结合第二方面或其上述实现方式的任一种，在第二方面的另一种实现方式中，所述第一组变换系数和所述第二组变换系数均按照相同方式划分为N个子组，所述确定单元具体用于确定N个目标值，其中，所述N个目标值中的第i目标值为所述第一组变换系数中的第i子组与所述第二组变换系数中的第i子组中各对应变换系数的差值之和；根据所述N个子组的加权系数，对所述N个目标值进行加权求和，得到所述图像差异值。
结合第二方面或其上述实现方式的任一种，在第二方面的另一种实现方式中，所述确定单元具体用于确定第一目标值，其中，所述第一目标值为所述第一组变换系数与所述第二组变换系数中各对应变换系数的差值之和；确定第二目标值，其中，所述第二目标值为所述第一变换系数集中除所述第一组变换系数外的剩余变换系数与所述第二变换系数集中除所述第二组变换系数外的剩余变换系数中各对应变换系数的差值之和；根据所述第一目标值和所述第二目标值的加权系数，对所述第一目标值和所述第二目标值进行加权求和，得到所述图像差异值。
结合第二方面或其上述实现方式的任一种，在第二方面的另一种实现方式中，所述第二变换系数集是对所述第一图像和所述第二图像的差值图像进行所述变换得到的变换系数集，所述确定单元具体用于根据所述第一变换系数集与所述第二变换系数集中变换系数的一一对应关系，从所述第二变换系数集中选取与所述第一组变换系数对应的第二组变换系数；根据所述第二组变换系数，确定所述图像差异值。
结合第二方面或其上述实现方式的任一种，在第二方面的另一种实现方 式中，所述确定单元具体用于将所述第二组变换系数中各变换系数之和确定为所述图像差异值。
结合第二方面或其上述实现方式的任一种，在第二方面的另一种实现方式中，所述第二组变换系数包括N个子组，所述确定单元具体用于确定N个目标值，其中，第i目标值为第i子组中的变换系数之和；根据所述N个目标值的加权系数，对所述N个目标值进行加权求和，得到所述图像差异值。
结合第二方面或其上述实现方式的任一种，在第二方面的另一种实现方式中，所述确定单元具体用于确定第一目标值，所述第一目标值为所述第二组变换系数中各变换系数之和；确定第二目标值，所述第二目标值为所述第二变换系数集中除所述第二组变换系数外的剩余变换系数之和；根据所述第一目标值和所述第二目标值的加权系数，对所述第一目标值和所述第二目标值进行加权求和，得到所述图像差异值。
结合第二方面或其上述实现方式的任一种，在第二方面的另一种实现方式中，所述第一图像为目标图像，所述第二图像为K个候选图像中的任意候选图像，所述处理单元具体用于根据所述目标图像和每个候选图像的图像差异值，确定所述每个候选图像的加权系数；根据所述K个候选图像的像素值，以及所述K个候选图像的加权系数，确定所述目标图像的滤波图像。
结合第二方面或其上述实现方式的任一种，在第二方面的另一种实现方式中，所述确定单元具体用于根据
Figure PCTCN2016070228-appb-000004
确定所述每个候选图像的加权系数，其中，b1、a1和h1均表示正实数，Dk表示所述第一图像和第k个候选图像的图像差异值，Wk表示所述第k个候选图像的加权系数。
结合第二方面或其上述实现方式的任一种，在第二方面的另一种实现方式中，所述确定单元具体用于根据Wk＝b2-(Dk)a2/h2确定所述每个候选图像的加权系数，其中，b2、a2和h2均表示正实数，Dk表示所述第一图像和第k个候选图像的图像差异值，Wk表示所述第k个候选图像的加权系数。
结合第二方面或其上述实现方式的任一种，在第二方面的另一种实现方 式中，所述确定单元具体用于根据
Figure PCTCN2016070228-appb-000005
确定所述每个候选图像的加权系数，其中，b3、a3和h3均表示正实数，Dk表示所述第一图像和第k个候选图像的图像差异值，Wk表示所述第k个候选图像的加权系数。
结合第二方面或其上述实现方式的任一种，在第二方面的另一种实现方式中，所述处理单元具体用于根据
Figure PCTCN2016070228-appb-000006
确定所述滤波图像，其中，Pfj表示所述滤波图像在第j像素点的像素值，W0表示所述第一图像的加权系数，Wk表示第k个候选图像的加权系数，P0j表示所述第一图像在第j像素点的像素值，Pkj表示所述第k个候选图像在第j像素点的像素值。
第三方面，提供一种解码方法，包括：从知识库中选取N个候选图像，其中，所述N个候选图像用于确定待解码图像的像素预测值，且每个候选图像与所述待解码图像的形状和大小均相同，所述待解码图像包括至少一个待解码的图像块；对所述待解码图像和N个目标图像中每个目标图像进行相同的变换，得到待解码图像的变换系数集和所述每个目标图像的变换系数集，所述N个目标图像为所述N个候选图像，或者所述N个目标图像为所述N个候选图像分别与所述待解码图像在对应像素点求差得到的图像，所述待解码图像的变换系数集与所述每个目标图像的变换系数集中的变换系数一一对应；从所述待解码图像的变换系数集中选取幅度满足预设阈值的变换系数，得到第一组变换系数；根据所述第一组变换系数，从所述每个目标图像的变换系数集中选出与所述第一组变换系数对应的第二组变换系数；根据所述N个目标图像对应的第二组变换系数，确定所述待解码图像与所述N个候选图像的图像差异值；根据所述待解码图像与所述N个候选图像的图像差异值，确定所述待解码图像的像素预测值；根据所述像素预测值对所述待解码图像进行解码。
结合第三方面，在第三方面的一种实现方式中，所述N个目标图像为所 述N个候选图像，所述根据所述N个目标图像对应的第二组变换系数，确定所述待解码图像与所述N个候选图像的图像差异值，包括：根据所述第一组变换系数和每个候选图像对应的第二组变换系数，确定所述待解码图像与所述每个候选图像的图像差异值。
结合第三方面或其上述实现方式的任一种，在第三方面的另一种实现方式中，所述根据所述第一组变换系数和每个候选图像对应的第二组变换系数，确定所述待解码图像与所述每个候选图像的图像差异值，包括：确定所述第一组变换系数和所述第二组变换系数中的各对应变换系数的差值；将所述各对应变换系数的差值求和，得到所述图像差异值。
结合第三方面或其上述实现方式的任一种，在第三方面的另一种实现方式中，所述根据所述第一组变换系数和每个候选图像对应的第二组变换系数，确定所述待解码图像与所述每个候选图像的图像差异值，包括：确定N个目标值，其中，所述第一组变换系数和所述第二组变换系数按照相同方式划分成N个子组，所述N个目标值中的第i目标值为所述第一组变换系数中的第i子组与所述第二组变换系数中的第i子组中各对应的变换系数的差值之和；根据所述N个目标值各自的加权系数，对所述N个目标值进行加权求和，得到所述图像差异值。
结合第三方面或其上述实现方式的任一种，在第三方面的另一种实现方式中，所述根据所述第一组变换系数和每个候选图像对应的第二组变换系数，确定所述待解码图像与所述每个候选图像的图像差异值，包括：确定第一目标值，其中，所述第一目标值为所述第一组变换系数与所述第二组变换系数中各对应变换系数的差值之和；确定第二目标值，其中，所述第二目标值为所述待解码图像的变换系数集中除所述第一组变换系数外的剩余变换系数与所述每个候选图像的变换系数集中除所述第二组变换系数外的剩余变换系数中各对应变换系数的差值之和；根据所述第一目标值和所述第二目标值的加权系数，对所述第一目标值和所述第二目标值进行加权求和，得到所述 图像差异值。
结合第三方面或其上述实现方式的任一种，在第三方面的另一种实现方式中，所述N个目标图像为所述N个候选图像分别与所述待解码图像在对应像素点求差得到的图像，所述根据所述N个目标图像对应的第二组变换系数，确定所述待解码图像与所述N个候选图像的图像差异值，包括：根据所述N个目标图像中每个目标图像对应的第二组变换系数，确定所述待解码图像与所述每个目标图像对应的候选图像的图像差异值，其中，所述每个目标图像对应的候选图像与所述待解码图像在对应像素点求差得到所述每个目标图像。
结合第三方面或其上述实现方式的任一种，在第三方面的另一种实现方式中，所述根据所述N个目标图像中每个目标图像对应的第二组变换系数，确定所述待解码图像与所述每个目标图像对应的候选图像的图像差异值，包括：将所述第二组变换系数中各变换系数之和确定为所述图像差异值。
结合第三方面或其上述实现方式的任一种，在第三方面的另一种实现方式中，所述根据所述N个目标图像中每个目标图像对应的第二组变换系数，确定所述待解码图像与所述每个目标图像对应的候选图像的图像差异值，包括：确定N个目标值，其中，所述第二组变换系数被划分成N个子组，所述N个目标值中的第i目标值为第i子组中的各变换系数之和；根据所述N个目标值的加权系数，对所述N个目标值进行加权求和，得到所述图像差异值。
结合第三方面或其上述实现方式的任一种，在第三方面的另一种实现方式中，所述根据所述N个目标图像中每个目标图像对应的第二组变换系数，确定所述待解码图像与所述每个目标图像对应的候选图像的图像差异值，包括：确定第一目标值，所述第一目标值为所述第二组变换系数中各显著变换系数之和；确定第二目标值，所述第二目标值为所述每个图像的变换系数集中除所述第二组变换系数外的剩余变换系数之和；根据所述第一目标值和所 述第二目标值的加权系数，对所述第一目标值和所述第二目标值进行加权求和，得到所述图像差异值。
结合第三方面或其上述实现方式的任一种，在第三方面的另一种实现方式中，所述根据所述待解码图像与所述多个候选图像的图像差异值，确定所述待解码图像的像素预测值，包括：根据所述待解码图像与所述多个候选图像的图像差异值，从所述多个候选图像中选出图像差异值最小的一个候选图像；将所述差异值最小的候选图像的像素值确定为所述待解码图像的像素预测值。
结合第三方面或其上述实现方式的任一种，在第三方面的另一种实现方式中，所述根据所述待解码图像与所述多个候选图像的图像差异值，确定所述待解码图像的像素预测值，包括：根据所述待解码图像与所述多个候选图像的图像差异值，从所述多个候选图像中选出图像差异值最小的E个候选图像，E≥2；根据所述待解码图像与所述E个候选图像中每个候选图像的图像差异值，确定所述每个候选图像的权重；根据所述E个候选图像的权重，对所述E候选图像的像素值进行加权平均，得到所述待解码图像的像素预测值。
结合第三方面或其上述实现方式的任一种，在第三方面的另一种实现方式中，所述从知识库中选取N个候选图像，包括：从码流中获取所述N个候选图像在所述知识库中的索引；根据所述索引，从所述知识库中选出所述N个候选图像。
第四方面，提供一种编码方法，包括：从知识库中选取N个候选图像，其中，所述N个候选图像用于确定待编码图像的像素预测值，且每个候选图像与所述待编码图像的形状和大小均相同，所述待编码图像包括至少一个待编码的图像块；对所述待编码图像和N个目标图像中每个目标图像进行相同的变换，得到待编码图像的变换系数集和所述每个目标图像的变换系数集，所述N个目标图像为所述N个候选图像，或者所述N个目标图像为所述N个候选图像分别与所述待编码图像在对应像素点求差得到的图像，所述待编 码图像的变换系数集与所述每个目标图像的变换系数集中的变换系数一一对应；从所述待编码图像的变换系数集中选取幅度满足预设阈值的变换系数，得到第一组变换系数；根据所述第一组变换系数，从所述每个目标图像的变换系数集中选出与所述第一组变换系数对应的第二组变换系数；根据所述N个目标图像对应的第二组变换系数，确定所述待编码图像与所述N个候选图像的图像差异值；根据所述待编码图像与所述N个候选图像的图像差异值，确定所述待编码图像的像素预测值；根据所述像素预测值对所述待编码图像进行编码。
结合第四方面，在第四方面的一种实现方式中，所述N个目标图像为所述N个候选图像，所述根据所述N个目标图像对应的第二组变换系数，确定所述待编码图像与所述N个候选图像的图像差异值，包括：根据所述第一组变换系数和每个候选图像对应的第二组变换系数，确定所述待编码图像与所述每个候选图像的图像差异值。
结合第四方面或其上述实现方式的任一种，在第四方面的另一种实现方式中，所述根据所述第一组变换系数和每个候选图像对应的第二组变换系数，确定所述待编码图像与所述每个候选图像的图像差异值，包括：确定所述第一组变换系数和所述第二组变换系数中的各对应变换系数的差值；将所述各对应变换系数的差值求和，得到所述图像差异值。
结合第四方面或其上述实现方式的任一种，在第四方面的另一种实现方式中，所述根据所述第一组变换系数和每个候选图像对应的第二组变换系数，确定所述待编码图像与所述每个候选图像的图像差异值，包括：确定N个目标值，其中，所述第一组变换系数和所述第二组变换系数按照相同方式划分成N个子组，所述N个目标值中的第i目标值为所述第一组变换系数中的第i子组与所述第二组变换系数中的第i子组中各对应的变换系数的差值之和；根据所述N个目标值各自的加权系数，对所述N个目标值进行加权求和，得到所述图像差异值。
结合第四方面或其上述实现方式的任一种，在第四方面的另一种实现方式中，所述根据所述第一组变换系数和每个候选图像对应的第二组变换系数，确定所述待编码图像与所述每个候选图像的图像差异值，包括：确定第一目标值，其中，所述第一目标值为所述第一组变换系数与所述第二组变换系数中各对应变换系数的差值之和；确定第二目标值，其中，所述第二目标值为所述待编码图像的变换系数集中除所述第一组变换系数外的剩余变换系数与所述每个候选图像的变换系数集中除所述第二组变换系数外的剩余变换系数中各对应变换系数的差值之和；根据所述第一目标值和所述第二目标值的加权系数，对所述第一目标值和所述第二目标值进行加权求和，得到所述图像差异值。
结合第四方面或其上述实现方式的任一种，在第四方面的另一种实现方式中，所述N个目标图像为所述N个候选图像分别与所述待编码图像在对应像素点求差得到的图像，所述根据所述N个目标图像对应的第二组变换系数，确定所述待编码图像与所述N个候选图像的图像差异值，包括：根据所述N个目标图像中每个目标图像对应的第二组变换系数，确定所述待编码图像与所述每个目标图像对应的候选图像的图像差异值，其中，所述每个目标图像对应的候选图像与所述待编码图像在对应像素点求差得到所述每个目标图像。
结合第四方面或其上述实现方式的任一种，在第四方面的另一种实现方式中，所述根据所述N个目标图像中每个目标图像对应的第二组变换系数，确定所述待编码图像与所述每个目标图像对应的候选图像的图像差异值，包括：将所述第二组变换系数中各变换系数之和确定为所述图像差异值。
结合第四方面或其上述实现方式的任一种，在第四方面的另一种实现方式中，所述根据所述N个目标图像中每个目标图像对应的第二组变换系数，确定所述待编码图像与所述每个目标图像对应的候选图像的图像差异值，包括：确定N个目标值，其中，所述第二组变换系数被划分成N个子组，所 述N个目标值中的第i目标值为第i子组中的各变换系数之和；根据所述N个目标值的加权系数，对所述N个目标值进行加权求和，得到所述图像差异值。
结合第四方面或其上述实现方式的任一种，在第四方面的另一种实现方式中，所述根据所述N个目标图像中每个目标图像对应的第二组变换系数，确定所述待编码图像与所述每个目标图像对应的候选图像的图像差异值，包括：确定第一目标值，所述第一目标值为所述第二组变换系数中各显著变换系数之和；确定第二目标值，所述第二目标值为所述每个图像的变换系数集中除所述第二组变换系数外的剩余变换系数之和；根据所述第一目标值和所述第二目标值的加权系数，对所述第一目标值和所述第二目标值进行加权求和，得到所述图像差异值。
结合第四方面或其上述实现方式的任一种，在第四方面的另一种实现方式中，所述根据所述待编码图像与所述多个候选图像的图像差异值，确定所述待编码图像的像素预测值，包括：根据所述待编码图像与所述多个候选图像的图像差异值，从所述多个候选图像中选出图像差异值最小的一个候选图像；将所述差异值最小的候选图像的像素值确定为所述待编码图像的像素预测值。
结合第四方面或其上述实现方式的任一种，在第四方面的另一种实现方式中，所述根据所述待编码图像与所述多个候选图像的图像差异值，确定所述待编码图像的像素预测值，包括：根据所述待编码图像与所述多个候选图像的图像差异值，从所述多个候选图像中选出图像差异值最小的E个候选图像，E≥2；根据所述待编码图像与所述E个候选图像中每个候选图像的图像差异值，确定所述每个候选图像的权重；根据所述E个候选图像的权重，对所述E候选图像的像素值进行加权平均，得到所述待编码图像的像素预测值。
结合第四方面或其上述实现方式的任一种，在第四方面的另一种实现方式中，所述方法还包括：将所述N个候选图像在所述知识库中的索引写入码 流。
第五方面，提供一种解码器，包括：
第一选取单元，用于从知识库中选取N个候选图像，其中，所述N个候选图像用于确定待解码图像的像素预测值，且每个候选图像与所述待解码图像的形状和大小均相同，所述待解码图像包括至少一个待解码的图像块；
变换单元，用于对所述待解码图像和N个目标图像中每个目标图像进行相同的变换，得到待解码图像的变换系数集和所述每个目标图像的变换系数集，所述N个目标图像为所述N个候选图像，或者所述N个目标图像为所述N个候选图像分别与所述待解码图像在对应像素点求差得到的图像，所述待解码图像的变换系数集与所述每个目标图像的变换系数集中的变换系数一一对应；
第二选取单元，用于从所述待解码图像的变换系数集中选取幅度满足预设阈值的变换系数，得到第一组变换系数；
第三选取单元，用于根据所述第一组变换系数，从所述每个目标图像的变换系数集中选出与所述第一组变换系数对应的第二组变换系数；
第一确定单元，用于根据所述N个目标图像对应的第二组变换系数，确定所述待解码图像与所述N个候选图像的图像差异值；
第二确定单元，用于根据所述待解码图像与所述N个候选图像的图像差异值，确定所述待解码图像的像素预测值；
解码单元，用于根据所述像素预测值对所述待解码图像进行解码。
结合第五方面，在第五方面的一种实现方式中，所述N个目标图像为所述N个候选图像，所述第一确定单元具体用于根据所述第一组变换系数和每个候选图像对应的第二组变换系数，确定所述待解码图像与所述每个候选图像的图像差异值。
结合第五方面或其上述实现方式的任一种，在第五方面的另一种实现方式中，所述第一确定单元具体用于确定所述第一组变换系数和所述第二组变 换系数中的各对应变换系数的差值；将所述各对应变换系数的差值求和，得到所述图像差异值。
结合第五方面或其上述实现方式的任一种，在第五方面的另一种实现方式中，所述第一确定单元具体用于确定N个目标值，其中，所述第一组变换系数和所述第二组变换系数按照相同方式划分成N个子组，所述N个目标值中的第i目标值为所述第一组变换系数中的第i子组与所述第二组变换系数中的第i子组中各对应的变换系数的差值之和；根据所述N个目标值各自的加权系数，对所述N个目标值进行加权求和，得到所述图像差异值。
结合第五方面或其上述实现方式的任一种，在第五方面的另一种实现方式中，所述第一确定单元具体用于确定第一目标值，其中，所述第一目标值为所述第一组变换系数与所述第二组变换系数中各对应变换系数的差值之和；确定第二目标值，其中，所述第二目标值为所述待解码图像的变换系数集中除所述第一组变换系数外的剩余变换系数与所述每个候选图像的变换系数集中除所述第二组变换系数外的剩余变换系数中各对应变换系数的差值之和；根据所述第一目标值和所述第二目标值的加权系数，对所述第一目标值和所述第二目标值进行加权求和，得到所述图像差异值。
结合第五方面或其上述实现方式的任一种，在第五方面的另一种实现方式中，所述N个目标图像为所述N个候选图像分别与所述待解码图像在对应像素点求差得到的图像，所述第一确定单元具体用于根据所述N个目标图像中每个目标图像对应的第二组变换系数，确定所述待解码图像与所述每个目标图像对应的候选图像的图像差异值，其中，所述每个目标图像对应的候选图像与所述待解码图像在对应像素点求差得到所述每个目标图像。
结合第五方面或其上述实现方式的任一种，在第五方面的另一种实现方式中，所述第一确定单元具体用于将所述第二组变换系数中各变换系数之和确定为所述图像差异值。
结合第五方面或其上述实现方式的任一种，在第五方面的另一种实现方 式中，所述第一确定单元具体用于确定N个目标值，其中，所述第二组变换系数被划分成N个子组，所述N个目标值中的第i目标值为第i子组中的各变换系数之和；根据所述N个目标值的加权系数，对所述N个目标值进行加权求和，得到所述图像差异值。
结合第五方面或其上述实现方式的任一种，在第五方面的另一种实现方式中，所述第一确定单元具体用于确定第一目标值，所述第一目标值为所述第二组变换系数中各显著变换系数之和；确定第二目标值，所述第二目标值为所述每个图像的变换系数集中除所述第二组变换系数外的剩余变换系数之和；根据所述第一目标值和所述第二目标值的加权系数，对所述第一目标值和所述第二目标值进行加权求和，得到所述图像差异值。
结合第五方面或其上述实现方式的任一种，在第五方面的另一种实现方式中，所述第二确定单元具体用于根据所述待解码图像与所述多个候选图像的图像差异值，从所述多个候选图像中选出图像差异值最小的一个候选图像；将所述差异值最小的候选图像的像素值确定为所述待解码图像的像素预测值。
结合第五方面或其上述实现方式的任一种，在第五方面的另一种实现方式中，所述第二确定单元具体用于根据所述待解码图像与所述多个候选图像的图像差异值，从所述多个候选图像中选出图像差异值最小的E个候选图像，E≥2；根据所述待解码图像与所述E个候选图像中每个候选图像的图像差异值，确定所述每个候选图像的权重；根据所述E个候选图像的权重，对所述E候选图像的像素值进行加权平均，得到所述待解码图像的像素预测值。
结合第五方面或其上述实现方式的任一种，在第五方面的另一种实现方式中，所述第一选取单元具体用于从码流中获取所述N个候选图像在所述知识库中的索引；根据所述索引，从所述知识库中选出所述N个候选图像。
第六方面，提供一种编码器，包括：
第一选取单元，用于从知识库中选取N个候选图像，其中，所述N个 候选图像用于确定待编码图像的像素预测值，且每个候选图像与所述待编码图像的形状和大小均相同，所述待编码图像包括至少一个待编码的图像块；
变换单元，用于对所述待编码图像和N个目标图像中每个目标图像进行相同的变换，得到待编码图像的变换系数集和所述每个目标图像的变换系数集，所述N个目标图像为所述N个候选图像，或者所述N个目标图像为所述N个候选图像分别与所述待编码图像在对应像素点求差得到的图像，所述待编码图像的变换系数集与所述每个目标图像的变换系数集中的变换系数一一对应；
第二选取单元，用于从所述待编码图像的变换系数集中选取幅度满足预设阈值的变换系数，得到第一组变换系数；
第三选取单元，用于根据所述第一组变换系数，从所述每个目标图像的变换系数集中选出与所述第一组变换系数对应的第二组变换系数；
第一确定单元，用于根据所述N个目标图像对应的第二组变换系数，确定所述待编码图像与所述N个候选图像的图像差异值；
第二确定单元，用于根据所述待编码图像与所述N个候选图像的图像差异值，确定所述待编码图像的像素预测值；
编码单元，用于根据所述像素预测值对所述待编码图像进行编码。
结合第六方面，在第六方面的一种实现方式中，所述N个目标图像为所述N个候选图像，所述第一确定单元具体用于根据所述第一组变换系数和每个候选图像对应的第二组变换系数，确定所述待编码图像与所述每个候选图像的图像差异值。
结合第六方面或其上述实现方式的任一种，在第六方面的另一种实现方式中，所述第一确定单元具体用于确定所述第一组变换系数和所述第二组变换系数中的各对应变换系数的差值；将所述各对应变换系数的差值求和，得到所述图像差异值。
结合第六方面或其上述实现方式的任一种，在第六方面的另一种实现方 式中，所述第一确定单元具体用于确定N个目标值，其中，所述第一组变换系数和所述第二组变换系数按照相同方式划分成N个子组，所述N个目标值中的第i目标值为所述第一组变换系数中的第i子组与所述第二组变换系数中的第i子组中各对应的变换系数的差值之和；根据所述N个目标值各自的加权系数，对所述N个目标值进行加权求和，得到所述图像差异值。
结合第六方面或其上述实现方式的任一种，在第六方面的另一种实现方式中，所述第一确定单元具体用于确定第一目标值，其中，所述第一目标值为所述第一组变换系数与所述第二组变换系数中各对应变换系数的差值之和；确定第二目标值，其中，所述第二目标值为所述待编码图像的变换系数集中除所述第一组变换系数外的剩余变换系数与所述每个候选图像的变换系数集中除所述第二组变换系数外的剩余变换系数中各对应变换系数的差值之和；根据所述第一目标值和所述第二目标值的加权系数，对所述第一目标值和所述第二目标值进行加权求和，得到所述图像差异值。
结合第六方面或其上述实现方式的任一种，在第六方面的另一种实现方式中，所述N个目标图像为所述N个候选图像分别与所述待编码图像在对应像素点求差得到的图像，所述第一确定单元具体用于根据所述N个目标图像中每个目标图像对应的第二组变换系数，确定所述待编码图像与所述每个目标图像对应的候选图像的图像差异值，其中，所述每个目标图像对应的候选图像与所述待编码图像在对应像素点求差得到所述每个目标图像。
结合第六方面或其上述实现方式的任一种，在第六方面的另一种实现方式中，所述第一确定单元具体用于将所述第二组变换系数中各变换系数之和确定为所述图像差异值。
结合第六方面或其上述实现方式的任一种，在第六方面的另一种实现方式中，所述第一确定单元具体用于确定N个目标值，其中，所述第二组变换系数被划分成N个子组，所述N个目标值中的第i目标值为第i子组中的各变换系数之和；根据所述N个目标值的加权系数，对所述N个目标值进行 加权求和，得到所述图像差异值。
结合第六方面或其上述实现方式的任一种，在第六方面的另一种实现方式中，所述第一确定单元具体用于确定第一目标值，所述第一目标值为所述第二组变换系数中各显著变换系数之和；确定第二目标值，所述第二目标值为所述每个图像的变换系数集中除所述第二组变换系数外的剩余变换系数之和；根据所述第一目标值和所述第二目标值的加权系数，对所述第一目标值和所述第二目标值进行加权求和，得到所述图像差异值。
结合第六方面或其上述实现方式的任一种，在第六方面的另一种实现方式中，所述第二确定单元具体用于根据所述待编码图像与所述多个候选图像的图像差异值，从所述多个候选图像中选出图像差异值最小的一个候选图像；将所述差异值最小的候选图像的像素值确定为所述待编码图像的像素预测值。
结合第六方面或其上述实现方式的任一种，在第六方面的另一种实现方式中，所述第二确定单元具体用于根据所述待编码图像与所述多个候选图像的图像差异值，从所述多个候选图像中选出图像差异值最小的E个候选图像，E≥2；根据所述待编码图像与所述E个候选图像中每个候选图像的图像差异值，确定所述每个候选图像的权重；根据所述E个候选图像的权重，对所述E候选图像的像素值进行加权平均，得到所述待编码图像的像素预测值。
结合第六方面或其上述实现方式的任一种，在第六方面的另一种实现方式中，所述编码器还包括：写入单元，用于将所述N个候选图像在所述知识库中的索引写入码流。
第七方面，提供一种解码器，用于从知识库中选取N个候选图像，其中，所述N个候选图像用于确定待解码图像的像素预测值，且每个候选图像与所述待解码图像的形状和大小均相同，所述待解码图像包括至少一个待解码的图像块；对所述待解码图像和N个目标图像中每个目标图像进行相同的变换，得到待解码图像的变换系数集和所述每个目标图像的变换系数集，所述N个 目标图像为所述N个候选图像，或者所述N个目标图像为所述N个候选图像分别与所述待解码图像在对应像素点求差得到的图像，所述待解码图像的变换系数集与所述每个目标图像的变换系数集中的变换系数一一对应；从所述待解码图像的变换系数集中选取幅度满足预设阈值的变换系数，得到第一组变换系数；根据所述第一组变换系数，从所述每个目标图像的变换系数集中选出与所述第一组变换系数对应的第二组变换系数；根据所述N个目标图像对应的第二组变换系数，确定所述待解码图像与所述N个候选图像的图像差异值；根据所述待解码图像与所述N个候选图像的图像差异值，确定所述待解码图像的像素预测值；根据所述像素预测值对所述待解码图像进行解码。
结合第七方面，在第七方面的一种实现方式中，所述N个目标图像为所述N个候选图像，所述编码器具体用于根据所述第一组变换系数和每个候选图像对应的第二组变换系数，确定所述待解码图像与所述每个候选图像的图像差异值。
结合第七方面或其上述实现方式的任一种，在第七方面的另一种实现方式中，所述编码器具体用于确定所述第一组变换系数和所述第二组变换系数中的各对应变换系数的差值；将所述各对应变换系数的差值求和，得到所述图像差异值。
结合第七方面或其上述实现方式的任一种，在第七方面的另一种实现方式中，所述编码器具体用于确定N个目标值，其中，所述第一组变换系数和所述第二组变换系数按照相同方式划分成N个子组，所述N个目标值中的第i目标值为所述第一组变换系数中的第i子组与所述第二组变换系数中的第i子组中各对应的变换系数的差值之和；根据所述N个目标值各自的加权系数，对所述N个目标值进行加权求和，得到所述图像差异值。
结合第七方面或其上述实现方式的任一种，在第七方面的另一种实现方式中，所述编码器具体用于确定第一目标值，其中，所述第一目标值为所述第一组变换系数与所述第二组变换系数中各对应变换系数的差值之和；确定 第二目标值，其中，所述第二目标值为所述待解码图像的变换系数集中除所述第一组变换系数外的剩余变换系数与所述每个候选图像的变换系数集中除所述第二组变换系数外的剩余变换系数中各对应变换系数的差值之和；根据所述第一目标值和所述第二目标值的加权系数，对所述第一目标值和所述第二目标值进行加权求和，得到所述图像差异值。
结合第七方面或其上述实现方式的任一种，在第七方面的另一种实现方式中，所述N个目标图像为所述N个候选图像分别与所述待解码图像在对应像素点求差得到的图像，所述编码器具体用于根据所述N个目标图像中每个目标图像对应的第二组变换系数，确定所述待解码图像与所述每个目标图像对应的候选图像的图像差异值，其中，所述每个目标图像对应的候选图像与所述待解码图像在对应像素点求差得到所述每个目标图像。
结合第七方面或其上述实现方式的任一种，在第七方面的另一种实现方式中，所述编码器具体用于将所述第二组变换系数中各变换系数之和确定为所述图像差异值。
结合第七方面或其上述实现方式的任一种，在第七方面的另一种实现方式中，所述编码器具体用于确定N个目标值，其中，所述第二组变换系数被划分成N个子组，所述N个目标值中的第i目标值为第i子组中的各变换系数之和；根据所述N个目标值的加权系数，对所述N个目标值进行加权求和，得到所述图像差异值。
结合第七方面或其上述实现方式的任一种，在第七方面的另一种实现方式中，所述编码器具体用于确定第一目标值，所述第一目标值为所述第二组变换系数中各显著变换系数之和；确定第二目标值，所述第二目标值为所述每个图像的变换系数集中除所述第二组变换系数外的剩余变换系数之和；根据所述第一目标值和所述第二目标值的加权系数，对所述第一目标值和所述第二目标值进行加权求和，得到所述图像差异值。
结合第七方面或其上述实现方式的任一种，在第七方面的另一种实现方 式中，所述编码器具体用于根据所述待解码图像与所述多个候选图像的图像差异值，从所述多个候选图像中选出图像差异值最小的一个候选图像；将所述差异值最小的候选图像的像素值确定为所述待解码图像的像素预测值。
结合第七方面或其上述实现方式的任一种，在第七方面的另一种实现方式中，所述编码器具体用于根据所述待解码图像与所述多个候选图像的图像差异值，从所述多个候选图像中选出图像差异值最小的E个候选图像，E≥2；根据所述待解码图像与所述E个候选图像中每个候选图像的图像差异值，确定所述每个候选图像的权重；根据所述E个候选图像的权重，对所述E候选图像的像素值进行加权平均，得到所述待解码图像的像素预测值。
结合第七方面或其上述实现方式的任一种，在第七方面的另一种实现方式中，所述编码器具体用于从码流中获取所述N个候选图像在所述知识库中的索引；根据所述索引，从所述知识库中选出所述N个候选图像。
第八方面，提供一种编码器，用于从知识库中选取N个候选图像，其中，所述N个候选图像用于确定待编码图像的像素预测值，且每个候选图像与所述待编码图像的形状和大小均相同，所述待编码图像包括至少一个待编码的图像块；对所述待编码图像和N个目标图像中每个目标图像进行相同的变换，得到待编码图像的变换系数集和所述每个目标图像的变换系数集，所述N个目标图像为所述N个候选图像，或者所述N个目标图像为所述N个候选图像分别与所述待编码图像在对应像素点求差得到的图像，所述待编码图像的变换系数集与所述每个目标图像的变换系数集中的变换系数一一对应；从所述待编码图像的变换系数集中选取幅度满足预设阈值的变换系数，得到第一组变换系数；根据所述第一组变换系数，从所述每个目标图像的变换系数集中选出与所述第一组变换系数对应的第二组变换系数；根据所述N个目标图像对应的第二组变换系数，确定所述待编码图像与所述N个候选图像的图像差异值；根据所述待编码图像与所述N个候选图像的图像差异值，确定所述待编码图像的像素预测值；根据所述像素预测值对所述待编码图像进行编码。
结合第八方面，在第八方面的一种实现方式中，所述N个目标图像为所述N个候选图像，所述编码器具体用于根据所述第一组变换系数和每个候选图像对应的第二组变换系数，确定所述待编码图像与所述每个候选图像的图像差异值。
结合第八方面或其上述实现方式的任一种，在第八方面的另一种实现方式中，所述编码器具体用于确定所述第一组变换系数和所述第二组变换系数中的各对应变换系数的差值；将所述各对应变换系数的差值求和，得到所述图像差异值。
结合第八方面或其上述实现方式的任一种，在第八方面的另一种实现方式中，所述编码器具体用于确定N个目标值，其中，所述第一组变换系数和所述第二组变换系数按照相同方式划分成N个子组，所述N个目标值中的第i目标值为所述第一组变换系数中的第i子组与所述第二组变换系数中的第i子组中各对应的变换系数的差值之和；根据所述N个目标值各自的加权系数，对所述N个目标值进行加权求和，得到所述图像差异值。
结合第八方面或其上述实现方式的任一种，在第八方面的另一种实现方式中，所述编码器具体用于确定第一目标值，其中，所述第一目标值为所述第一组变换系数与所述第二组变换系数中各对应变换系数的差值之和；确定第二目标值，其中，所述第二目标值为所述待编码图像的变换系数集中除所述第一组变换系数外的剩余变换系数与所述每个候选图像的变换系数集中除所述第二组变换系数外的剩余变换系数中各对应变换系数的差值之和；根据所述第一目标值和所述第二目标值的加权系数，对所述第一目标值和所述第二目标值进行加权求和，得到所述图像差异值。
结合第八方面或其上述实现方式的任一种，在第八方面的另一种实现方式中，所述N个目标图像为所述N个候选图像分别与所述待编码图像在对应像素点求差得到的图像，所述编码器具体用于根据所述N个目标图像中每个目标图像对应的第二组变换系数，确定所述待编码图像与所述每个目标图 像对应的候选图像的图像差异值，其中，所述每个目标图像对应的候选图像与所述待编码图像在对应像素点求差得到所述每个目标图像。
结合第八方面或其上述实现方式的任一种，在第八方面的另一种实现方式中，所述编码器具体用于将所述第二组变换系数中各变换系数之和确定为所述图像差异值。
结合第八方面或其上述实现方式的任一种，在第八方面的另一种实现方式中，所述编码器具体用于确定N个目标值，其中，所述第二组变换系数被划分成N个子组，所述N个目标值中的第i目标值为第i子组中的各变换系数之和；根据所述N个目标值的加权系数，对所述N个目标值进行加权求和，得到所述图像差异值。
结合第八方面或其上述实现方式的任一种，在第八方面的另一种实现方式中，所述编码器具体用于确定第一目标值，所述第一目标值为所述第二组变换系数中各显著变换系数之和；确定第二目标值，所述第二目标值为所述每个图像的变换系数集中除所述第二组变换系数外的剩余变换系数之和；根据所述第一目标值和所述第二目标值的加权系数，对所述第一目标值和所述第二目标值进行加权求和，得到所述图像差异值。
结合第八方面或其上述实现方式的任一种，在第八方面的另一种实现方式中，所述编码器具体用于根据所述待编码图像与所述多个候选图像的图像差异值，从所述多个候选图像中选出图像差异值最小的一个候选图像；将所述差异值最小的候选图像的像素值确定为所述待编码图像的像素预测值。
结合第八方面或其上述实现方式的任一种，在第八方面的另一种实现方式中，所述编码器具体用于根据所述待编码图像与所述多个候选图像的图像差异值，从所述多个候选图像中选出图像差异值最小的E个候选图像，E≥2；根据所述待编码图像与所述E个候选图像中每个候选图像的图像差异值，确定所述每个候选图像的权重；根据所述E个候选图像的权重，对所述E候选图像的像素值进行加权平均，得到所述待编码图像的像素预测值。
结合第八方面或其上述实现方式的任一种，在第八方面的另一种实现方式中，所述编码器还用于将所述N个候选图像在所述知识库中的索引写入码流。
本发明实施例中，幅度满足预设的阈值条件的变换系数能够更好地反应图像的主要结构信息，利用幅度满足预设的阈值条件的变换系数确定出的图像差异值能够较好地反映图像间的差异程度，使得基于该图像差异值的后续图像处理更加准确。
附图说明
为了更清楚地说明本发明实施例的技术方案，下面将对本发明实施例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是本发明实施例的处理图像的方法的示意性流程图。
图2是预测块的目标模版图像位置关系的示例图。
图3是本发明实施例的处理图像的装置的示意性框图。
图4是本发明实施例的处理图像的装置的示意性框图。
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，所描述的实施例是本发明的一部分实施例，而不是全部实施例。基于本发明中的实施例，本领域普通技术人员在没有作出创造性劳动的前提下所获得的所有其他实施例，都应属于本发明保护的范围。
图1是本发明实施例的图像处理方法的示意性流程图。图1的方法包括：
110、对第一图像进行变换，得到第一变换系数集。
120、对第二图像进行变换，或者对第一图像和第二图像的差值图像进行变换，得到第二变换系数集。
应理解，本发明实施例中的图像也可以称为图样(patch)。第一图像和第二图像的差值图像具体可指第一图像和第二图像在对应像素点求差后得到的图像。
应理解，第一图像和第二图像可以是相同形状、相同大小的图像区域，包括相同数量的像素，其形状例如可以是矩形、L形、三角形、菱形、梯形、六边形等。
此外，第一图像可以是一个解码预测单元的预测图像，或者是解码预测单元对应的模版区域的重建图像，又或者是解码图像中一个预设形状的重建图像。第二图像可以是一个解码端知识库中的高质量图像，例如预先存储在知识库中的原始图像；第二图像也可以是一个已解码图像中的局部区域。
此外，对第一图像和第二图像分别进行相同的变换，得到第一变换系数集和第二变换系数集。其中，此处的变换可以采用多种常用的变换方式，例如采用离散余弦变换(Discrete Cosine Transform,DCT)、离散正弦变换(Discrete Sine Transform,DST)、哈达玛变换(Hadamard Transform)、小波变换(Wavelet Transform)、尺寸不变特征变换(Scale-invariant Feature Transform,SIFT)变换等。以DCT变换为例，图像为矩形图像时，可以用二维DCT变换将图像变换为一个二维变换矩阵，也可以将图像中的像素排列成一维矢量，用一维DCT变换将图像变换成一维矢量；又例如，当图像为三角形图像时，可以将图像中的像素排列成一维矢量，进行一维DCT变换，也可以将图像重新排列成矩形，进行二维DCT变换。
此外，变换系数集为图像经过变换得到的变换系数构成的集合，它可以包含图像经过变换得到的全部变换系数，也可以包含全部变换系数中的一部分系数，例如若干个预设系数位置的变换系数。例如，一个变换系数集可包含N个变换系数Ci(i＝1,2,...,N)，每个变换系数可对应一个编号(index)i，N为正整数。上述第一变换系数集和第二变换系数集可使用相同的变换系数编号方式。编号方式可以为多种形式，例如当使用二维DCT变换得到变换 系数矩阵时，可以按照ZigZag扫描顺序对变换系数进行编号，或者也可以按照从左到右、从上到下的顺序对变换系数进行编号。
在变换时，可以针对像素的一个或多个色彩空间分量(color space component)进行变换。例如，可以只对其中的一个色彩空间分量(例如亮度分量)进行变换，后续会得到一个图像差异值；也可以对多个色彩空间分量分别进行变换，后续会得到多个图像差异值，再将每个色彩空间分量下的图像差异值求平均值，得到一个最终的图像差异值；当然，还可以将所有色彩空间分量组合在一起，统一进行变换，后续会得到一个图像差异值。
130、根据第一变换系数集中的变换系数的幅度，从第一变换系数集中选取第一组变换系数，其中，第一组变换系数中的变换系数的幅度满足预设的阈值条件。
上述预设的阈值条件可以指：第一组变换系数中的各变换系数的幅度不小于所述第一变换系数集中除所述第一组变换系数之外的剩余变换系数的幅度或者所述第一组变换系数中任一变换系数的幅度不小于该变换系数对应的阈值；
举例说明，可以采用如下两种处理方式中的任一种从第一变换系数集中获取第一组变换系数：
处理方式一：判断第一变换系数集中每个变换系数Ci的幅度是否超过一个阈值THi；如果该系数的幅度越过该阈值，将该系数加入上述第一组变换系数中。每个变换系数对比的阈值THi可以相同，也可以不同。该阈值可以为预设的常数，也可以为第一图像对应量化步长的x倍(x为正实数)；当第一图像为解码图像中的图像时，上述阈值还可以从该解码图像对应的码流中提取得到。
处理方式二：选取第一变换系数集中幅度最大的M个系数作为显著变换系数。其中，M为正整数，例如M为4、8、15、16、20或32。
需要说明的是，上述第一组变换系数为第一变换系数集中的哪些变换系 数可以通过变换系数信息的方式进行指示。该变换系数信息可以有多种形式，例如变换系数信息可以为第一组变换系数在第一变换系数集中的编号；又如，变换系数信息可以为一个数组，该数组的维数与第一变换系数集中变换系数的个数相同，且数组中每一个元素指示第一变换系数集中的一个变换系数是否属于该第一组变换系数。
140、根据第一组变换系数和第二变换系数集，确定第一图像和第二图像之间的图像差异值。
具体而言，当第二变换系数是对第二图像进行变换得到的变换系数集时，步骤140可包括：根据第一组变换系数，以及第一变换系数集与第二变换系数集中变换系数的一一对应关系，从第二变换系数集中选出第二组变换系数；根据第一组变换系数和第二组变换系数的差值，确定第一图像和第二图像之间的图像差异值。当第二变换系数是对第一图像和第二图像的差值图像进行变换后得到的变换系数集时，步骤140可包括：根据第一组变换系数，以及第一变换系数集与第二变换系数集中变换系数的一一对应关系，从第二变换系数集中选出第二组变换系数；根据第二组变换系数，确定第一图像和第二图像之间的差异值。
150、根据图像差异值，对第一图像和第二图像进行处理。
上述处理可以包括以下处理中的至少一种：
1)分别确定第一图像和K个第二图像中每一个图像之间的图像差异值；根据K个图像差异值，确定加权系数，加权系数为关于图像差异值的减函数；根据加权系数，对第一图像和K个第二图像进行滤波处理，得到滤波图像；
2)分别确定第一图像和K个第二图像中每一个图像之间的图像差异值；将K个图像差异值中最小的G个第二图像，作为第一图像的最近邻图像；
3)分别确定第一图像和K个第二图像中每一个图像之间的图像差异值；根据K个图像差异值，确定加权系数，加权系数为关于图像差异值的减函数；根据加权系数，对K个第二图像的相邻图像进行加权平均，得到第一图像的 相邻图像的预测图像；(对应实施例5)
4)分别确定第一图像和K个第二图像中每一个图像之间的图像差异值；根据K个图像差异值，确定加权系数，加权系数为关于图像差异值的减函数；根据加权系数，对第一图像和K个第二图像进行滤波处理，得到滤波图像，将滤波图像作为第一图像所在图像区域的解码重建图像；
5)分别确定第一图像和S个第二图像中每一个图像之间的图像差异值；将S个图像差异值中最小的E个第二图像，作为第一图像的最近邻图像；对E个第二图像计算加权平均或者选择第二图像中图像差异值最小的图像，得到第一图像的预测图像；根据预测图像对第一图像进行编码处理。
本发明实施例中，幅度满足预设的阈值条件的变换系数能够更好地反应图像的主要结构信息，利用幅度满足预设的阈值条件的变换系数确定出的图像差异值能够较好地反映图像间的差异程度，使得基于该图像差异值的后续图像处理更加准确。
可选地，作为一个实施例，第二变换系数集是对第二图像进行变换得到的变换系数集，步骤140可包括：根据第一变换系数集与第二变换系数集中变换系数的一一对应关系，从第二变换系数集中选取与第一组变换系数对应的第二组变换系数；根据第一组变换系数和第二组变换系数，确定图像差异值。
可选地，作为一个实施例，上述根据第一组变换系数和第二组变换系数，确定图像差异值可包括：确定第一组变换系数和第二组变换系数中各对应变换系数的差值；将各对应变换系数的差值之和确定为图像差异值。
可选地，作为一个实施例，第一组变换系数和第二组变换系数均按照相同方式划分为N个子组，上述根据第一组变换系数和第二组变换系数，确定图像差异值可包括：确定N个目标值，其中，N个目标值中的第i目标值为第一组变换系数中的第i子组与第二组变换系数中的第i子组中各对应变换系数的差值之和；根据N个子组的加权系数，对N个目标值进行加权求和， 得到图像差异值。
应理解，上述加权系数(weight)为实数，通常为正实数，子组的加权系数可以由子组中的变换系数在变换系数集中的位置决定。例如，当第一图像的变换为DCT或DST变换时，将所有变换系数按照常用的ZigZag扫描顺序进行编号，第一变换系数集中的编号值小于总编号值1/2的显著变换系数(即变换系数的幅度满足预设的阈值条件的变换系数)划分为子组1，其加权系数为0.75；第一变换系数集中的其余显著变换系数划分子组2，其加权系数为0.25。又例如，当变换为DCT或DST变换且第一图像为矩形时，第一变换系数集中的位于变换系数矩阵左上角1/4区域中的显著变换系数划分为子组1，其加权系数为1.2；将第一变换系数集中的位于变换系数矩阵中右上角1/4区域和左下角1/4区域中的显著变换系数划分为子组2，其加权系数为0.8；将第一变换系数集中的位于变换系数矩阵中右下角1/4区域的显著变换系数划分为子组3，其加权系数为0.25。
可选地，作为一个实施例，上述根据第一组变换系数和第二组变换系数，确定图像差异值可包括：确定第一目标值，其中，第一目标值为第一组变换系数与第二组变换系数中各对应变换系数的差值之和；确定第二目标值，其中，第二目标值为第一变换系数集中除第一组变换系数外的剩余变换系数与第二变换系数集中除第二组变换系数外的剩余变换系数中各对应变换系数的差值之和；根据第一目标值和第二目标值的加权系数，对第一目标值和第二目标值进行加权求和，得到图像差异值。
具体地，第一目标值的加权系数可以为正实数，例如可以选取1、0.5或2等，第二目标值的加权系数可以为非零实数，例如可以选取-0.2、0.5或0.8等；第一目标值的加权系数和第二目标值的加权系数可以不相等，通常第一目标值的加权系数大于第二目标值的加权系数。
应理解，上文提到的差值之和可以是均方误差(MSE)、平均绝对差(MAD)、平方误差和(SSE)、绝对误差和(SAD)其中之一，也可以使用 其它的求和计算方法。更具体的，例如当差值之和使用平方误差和时，将两组变换系数中对应的变换系数两两相减得到一组差值，将每个差值的平方相加得到平方误差和；又例如当差值之和使用平均绝对差时，将两组变换系数中对应的变换系数相减得到一组差值，将每个差值的绝对值相加得到绝对误差和，再将绝对误差和除以差值的个数进行归一化，得到平均绝对差。
可选地，作为一个实施例，第二变换系数集是对第一图像和第二图像的差值图像进行变换得到的变换系数集，步骤140可包括：根据第一变换系数集与第二变换系数集中变换系数的一一对应关系，从第二变换系数集中选取与第一组变换系数对应的第二组变换系数；根据第二组变换系数，确定图像差异值。
可选地，作为一个实施例，上述根据第二组变换系数，确定图像差异值，包括：将第二组变换系数中各变换系数之和确定为图像差异值。
可选地，作为一个实施例，第二组变换系数包括N个子组，上述根据第二组变换系数，确定图像差异值可包括：确定N个目标值，其中，第i目标值为第i子组中的变换系数之和；根据N个目标值的加权系数，对N个目标值进行加权求和，得到图像差异值。
可选地，作为一个实施例，上述根据第二组变换系数，确定图像差异值可包括：确定第一目标值，第一目标值为第二组变换系数中各变换系数之和；确定第二目标值，第二目标值为第二变换系数集中除第二组变换系数外的剩余变换系数之和；根据第一目标值和第二目标值的加权系数，对第一目标值和第二目标值进行加权求和，得到图像差异值。
具体地，第一数值总和的加权系数可以为正实数，例如1、0.5或2等，第二数值总和的加权系数可以为非零实数，例如-0.2、0.5、0.8或0.9等；第一数值总和的加权系数和第二数值总和的加权系数可以不相等，通常第一数值总和的加权系数大于第二数值总和的加权系数。
需要说明的是，上述变换系数之和可以是绝对值之和、平方值之和、平 方值的均值、绝对值的均值、均方根(Root Mean Squared Error,RMSE)其中之一，也可以使用其它的计算形式。更具体的，例如当变换系数之和使用绝对值之和时，将每个系数的绝对值相加得到绝对值之和；又例如当变换系数之和使用平方值的均值时，将变换系数的平方值相加得到平方值之和，再将平方值之和除以这些变换系数的个数进行归一化，得到平方值的均值。
下面结合具体的实施例，详细描述利用上述图像差异值进行后续处理的多种方式。
可选地，作为一个实施例，第一图像为目标图像，第二图像为K个候选图像中的任意候选图像，步骤140可包括：根据目标图像和每个候选图像的图像差异值，确定所述每个候选图像的加权系数；根据K个候选图像的像素值，以及K个候选图像的加权系数，确定目标图像的滤波图像。
上述滤波图像的视觉质量高于目标图像，将滤波图像替换目标图像，可以增强目标图像所在图像的视觉质量。
具体地，上述K个候选图像中任意一个候选图像与目标图像可具有相同形状和大小。目标图像可以是一幅图像中一个区域，例如解码图像中一个预设形状的图像，预设形状例如为矩形或三角形。此外，目标图像与K个候选图像中每个候选图像具有一个图像差异值，且目标图像与每个候选图像之间图像差异值的确定方式均可采用上述实施例描述的方式。
上述像素值可以为像素的色彩空间分量之一，此时像素值为一个标量；像素值也可以是像素的色彩空间分量中多个分量组成的多维矢量。像素的色彩空间例如RGB、YCbCr、YUV、HSV、CIE Lab等常用色彩空间。
上述根据目标图像和每个候选图像之间的图像差异值，确定每个候选图像的加权系数的方式可以有多种实现方式，下面介绍几种实现方式。
例如，根据
Figure PCTCN2016070228-appb-000007
确定每个候选图像的加权系数，其中，b1、a1和h1均表示正实数，Dk表示目标图像和第k个候选图像的图像差异值，Wk表示第k候选图像的加权系数。
又如，根据Wk＝b2-(Dk)a2/h2确定每个候选图像对应的加权系数，其中，b2、a2和h2均表示正实数，Dk表示目标图像和第k个候选图像的图像差异值，Wk表示第k候选图像的加权系数。
又如，根据
Figure PCTCN2016070228-appb-000008
确定每个候选图像的加权系数，其中，b3、a3和h3均表示正实数，Dk表示目标图像和第k个候选图像的图像差异值，Wk表示第k候选图像的加权系数。
此外，上述根据K个候选图像的像素值，以及K个候选图像的加权系数，确定目标图像的滤波图像的方式也有多种，例如可以根据
Figure PCTCN2016070228-appb-000009
确定目标图像的滤波图像，其中，Pfj表示滤波图像在第j像素点的像素值，W0表示目标图像的加权系数，Wk表示第k候选图像的加权系数，P0j表示目标图像在第j像素点的像素值，Pkj表示第k候选图像在第j像素点的像素值。
可选地，作为一个实施例，第一图像为目标图像，第二图像为K个候选图像中的任意候选图像，根据图像差异值，对第一图像和第二图像进行后续图像处理，包括：根据目标图像与K个候选图像的图像差异值，从K个候选图像中选取与目标图像的图像差异值最小的G个候选图像，作为目标图像的最近邻图像，其中，G小于等于K。
本实施例可以看成是一种图像最近邻(nearest neighbor)搜索方法。目标图像可以是一幅图像中某个图像区域，例如解码图像中一个预设形状的图像，目标图像也可以是一个解码预测单元的预测图像，例如通过H.265中基于运动矢量的帧间预测得到的预测图像；预设形状例如为矩形或三角形。K个候选图像可以是当前解码图像已重建区域中的图像或者其它已解码图像中的图像，也可以是解码端知识库中的图像。
应理解，目标图像与K个候选图像中每个候选图像具有图像差异值，且 目标图像与每个候选图像的图像差异值的确定方式可以采用上述任意实施例中的图像差异值确定方式。
应理解，找到图像差异值最小的G个图像，作为目标图像的最近邻图像，其中，G为正整数，例如G＝1、2、4、6、8等。
需要说明的是，最近邻图像可以用于增强目标图像，例如将图像差异值最小的最近邻图像替换目标图像；又例如采用最近邻图像对目标图像进行滤波，将上述G个最近邻图像与目标图像一起合成一个滤波图像，来替换目标图像P0。最近邻图像还可以用于指示迭代搜索的起始位置，例如先在起始搜索点周围的若干个图像中找到最近邻图像，再以最近邻图像为新的搜索起始点，进行下一次迭代搜索。
可选地，作为一个实施例，第一图像为预测块对应的目标模板图像，其中，预测块对应K个候选块，K个候选块分别对应K个候选模板图像，第二图像为K个候选模板图像分别与目标模板图像在对应像素点求差得到的K个残差模板图像中的任意残差模板图像，根据图像差异值，对第一图像和第二图像进行后续图像处理，包括：根据目标模板图像与K个残差模板图像的图像差异值，从K个残差模板图像中选取与目标模板图像的图像差异值最小的G个残差模板图像，其中，G小于等于K；从K个候选块中选取G个残差模板图像对应的G个候选块；根据G个候选块的像素值，确定预测块的像素预测值。需要说明的是，也可以是第一图像为预测块对应的目标模板图像，第三图像为K个候选模板图像分别与目标模板图像在对应像素点求差得到的K个残差模板图像中的任意残差模板图像。
本实施例可以看成一种模版匹配预测方法。预测块通常为一个矩形区域，它和任意一个候选块可具有相同形状和大小；目标模版图像和任意一个候选模版图像可具有相同形状和大小；预测块和目标模版图像构成的区域与任意一对候选块和候选模版图像构成的区域可具有相同形状和大小。目标模版图像可与预测块邻接，通常为矩形区域或者L形区域，也可以是其它的形状， 图2给出了预测块B0的目标模版图像T0的几种示例。
此外，候选模版图像可以是解码端知识库中的图像，也可以是当前解码图像已重建区域中的图像或者其它已解码图像中的图像。
找到图像差异值最小的G个候选模版图像对应的候选块Bg(g＝1,2,..,G)，作为预测块B0的最近邻图像；其中，G为正整数，且G≤Q，例如G＝1、2、4、6、8等。
上述根据G个候选块的像素值，确定预测块的像素预测值可包括：根据
Figure PCTCN2016070228-appb-000010
确定预测块的像素预测值，其中，B0j表示预测块在第j像素点的像素值，Bgj表示第g候选块在第j像素点的像素值，Wg表示第g候选块对应的加权系数。
候选块对应的加权系数Wg可以根据该候选块的候选模版图像与目标模版图像的图像差异值Dg决定。例如，加权系数Wg随着Dg增大而减小，具体地，可以根据
Figure PCTCN2016070228-appb-000011
确定G个候选块对应的加权系数，其中，b1、a1和h1均表示正实数，Dg表示目标图像和第g个候选图像之间的图像差异值，Wg表示第g候选块对应的加权系数。
又如，根据Wg＝b2-(Dg)a2/h2确定G个候选图像中每个候选图像的加权系数，其中，b2、a2和h2均表示正实数，Dg表示目标图像和第g个候选图像的图像差异值，Wg表示第g候选图像的加权系数。
又如，根据
Figure PCTCN2016070228-appb-000012
确定G个候选图像中每个候选图像的加权系数，其中，b3、a3和h3均表示正实数，Dg表示目标图像和第g个候选图像的图像差异值，Wg表示第g候选图像的加权系数。
上文中结合图1至图2，详细描述了根据本发明实施例的处理图像的方法，下面将结合图3至图4，描述根据本发明实施例的处理图像的装置。
应理解，图3和图4描述的处理图像的装置能够实现图1中描述的处理图像的方法的各个步骤，为了简洁，适当省略重复的描述。
图3是本发明实施例的处理图像的装置的示意性框图。图3的装置300包括：
第一变换单元310，用于对第一图像进行变换，得到第一变换系数集；
第二变换单元320，用于对第二图像进行所述变换，或者对第一图像和第二图像的差值图像进行所述变换，得到第二变换系数集；
选取单元330，用于根据所述第一变换单元310得到的所述第一变换系数集中的变换系数的幅度，从所述第一变换系数集中选取第一组变换系数，其中，所述第一组变换系数中的变换系数的幅度满足预设的阈值条件；
确定单元340，用于根据所述选取单元330选取的所述第一组变换系数和所述第二变换单元320得到的所述第二变换系数集，确定第一图像和第二图像之间的图像差异值；
处理单元350，用于根据所述确定单元340确定的所述图像差异值，对所述第一图像和所述第二图像进行处理。
本发明实施例中，幅度满足预设的阈值条件的变换系数能够更好地反应图像的主要结构信息，利用幅度满足预设的阈值条件的变换系数确定出的图像差异值能够较好地反映图像间的差异程度，使得基于该图像差异值的后续图像处理更加准确。
可选地，作为一个实施例，所述第二变换系数集是对所述第二图像进行所述变换得到的变换系数集，所述确定单元340具体用于根据所述第一变换系数集与所述第二变换系数集中变换系数的一一对应关系，从所述第二变换系数集中选取与所述第一组变换系数对应的第二组变换系数；根据所述第一组变换系数和所述第二组变换系数，确定所述图像差异值。
可选地，作为一个实施例，所述确定单元340具体用于确定所述第一组变换系数和所述第二组变换系数中各对应变换系数的差值；将所述各对应变 换系数的差值之和确定为所述图像差异值。
可选地，作为一个实施例，所述第一组变换系数和所述第二组变换系数均按照相同方式划分为N个子组，所述确定单元340具体用于确定N个目标值，其中，所述N个目标值中的第i目标值为所述第一组变换系数中的第i子组与所述第二组变换系数中的第i子组中各对应变换系数的差值之和；根据所述N个子组的加权系数，对所述N个目标值进行加权求和，得到所述图像差异值。
可选地，作为一个实施例，所述确定单元340具体用于确定第一目标值，其中，所述第一目标值为所述第一组变换系数与所述第二组变换系数中各对应变换系数的差值之和；确定第二目标值，其中，所述第二目标值为所述第一变换系数集中除所述第一组变换系数外的剩余变换系数与所述第二变换系数集中除所述第二组变换系数外的剩余变换系数中各对应变换系数的差值之和；根据所述第一目标值和所述第二目标值的加权系数，对所述第一目标值和所述第二目标值进行加权求和，得到所述图像差异值。
可选地，作为一个实施例，所述第二变换系数集是对所述第一图像和所述第二图像的差值图像进行所述变换得到的变换系数集，所述确定单元340具体用于根据所述第一变换系数集与所述第二变换系数集中变换系数的一一对应关系，从所述第二变换系数集中选取与所述第一组变换系数对应的第二组变换系数；根据所述第二组变换系数，确定所述图像差异值。
可选地，作为一个实施例，所述确定单元340具体用于将所述第二组变换系数中各变换系数之和确定为所述图像差异值。
可选地，作为一个实施例，所述第二组变换系数包括N个子组，所述确定单元340具体用于确定N个目标值，其中，第i目标值为第i子组中的变换系数之和；根据所述N个目标值的加权系数，对所述N个目标值进行加权求和，得到所述图像差异值。
可选地，作为一个实施例，所述确定单元340具体用于确定第一目标值， 所述第一目标值为所述第二组变换系数中各变换系数之和；确定第二目标值，所述第二目标值为所述第二变换系数集中除所述第二组变换系数外的剩余变换系数之和；根据所述第一目标值和所述第二目标值的加权系数，对所述第一目标值和所述第二目标值进行加权求和，得到所述图像差异值。
可选地，作为一个实施例，所述第一图像为目标图像，所述第二图像为K个候选图像中的任意候选图像，所述处理单元350具体用于根据所述目标图像和每个候选图像的图像差异值，确定所述每个候选图像的加权系数；根据所述K个候选图像的像素值，以及所述K个候选图像的加权系数，确定所述目标图像的滤波图像。
可选地，作为一个实施例，所述确定单元340具体用于根据
Figure PCTCN2016070228-appb-000013
确定所述每个候选图像的加权系数，其中，b1、a1和h1均表示正实数，Dk表示所述第一图像和第k个候选图像的图像差异值，Wk表示所述第k个候选图像的加权系数。
可选地，作为一个实施例，所述确定单元340具体用于根据Wk＝b2-(Dk)a2/h2确定所述每个候选图像的加权系数，其中，b2、a2和h2均表示正实数，Dk表示所述第一图像和第k个候选图像的图像差异值，Wk表示所述第k个候选图像的加权系数。
可选地，作为一个实施例，所述确定单元340具体用于根据
Figure PCTCN2016070228-appb-000014
确定所述每个候选图像的加权系数，其中，b3、a3和h3均表示正实数，Dk表示所述第一图像和第k个候选图像的图像差异值，Wk表示所述第k个候选图像的加权系数。
可选地，作为一个实施例，所述处理单元350具体用于根据
Figure PCTCN2016070228-appb-000015
确定所述滤波图像，其中，Pfj表示所述滤波图像在第j像素点的像素值，W0表示所述第一图像的加权系数，Wk表示第k个候选 图像的加权系数，P0j表示所述第一图像在第j像素点的像素值，Pkj表示所述第k个候选图像在第j像素点的像素值。
图4是本发明实施例的处理图像的装置的示意性框图。图4的装置400包括：
存储器410，用于存储程序；
处理器420，用于执行程序，当所述程序被执行时，所述处理器420具体用于：对第一图像进行变换，得到第一变换系数集；对第二图像进行所述变换，或者对第一图像和第二图像的差值图像进行所述变换，得到第二变换系数集；根据所述第一变换系数集中的变换系数的幅度，从所述第一变换系数集中选取第一组变换系数，其中，所述第一组变换系数中的变换系数的幅度满足预设的阈值条件；根据所述第一组变换系数和所述第二变换系数集，确定第一图像和第二图像之间的图像差异值；根据所述图像差异值，对所述第一图像和所述第二图像进行处理。
本发明实施例中，幅度满足预设的阈值条件的变换系数能够更好地反应图像的主要结构信息，利用幅度满足预设的阈值条件的变换系数确定出的图像差异值能够较好地反映图像间的差异程度，使得基于该图像差异值的后续图像处理更加准确。
可选地，作为一个实施例，所述第二变换系数集是对所述第二图像进行所述变换得到的变换系数集，所述处理器420具体用于根据所述第一变换系数集与所述第二变换系数集中变换系数的一一对应关系，从所述第二变换系数集中选取与所述第一组变换系数对应的第二组变换系数；根据所述第一组变换系数和所述第二组变换系数，确定所述图像差异值。
可选地，作为一个实施例，所述处理器420具体用于确定所述第一组变换系数和所述第二组变换系数中各对应变换系数的差值；将所述各对应变换系数的差值之和确定为所述图像差异值。
可选地，作为一个实施例，所述第一组变换系数和所述第二组变换系数 均按照相同方式划分为N个子组，所述处理器420具体用于确定N个目标值，其中，所述N个目标值中的第i目标值为所述第一组变换系数中的第i子组与所述第二组变换系数中的第i子组中各对应变换系数的差值之和；根据所述N个子组的加权系数，对所述N个目标值进行加权求和，得到所述图像差异值。
可选地，作为一个实施例，所述处理器420具体用于确定第一目标值，其中，所述第一目标值为所述第一组变换系数与所述第二组变换系数中各对应变换系数的差值之和；确定第二目标值，其中，所述第二目标值为所述第一变换系数集中除所述第一组变换系数外的剩余变换系数与所述第二变换系数集中除所述第二组变换系数外的剩余变换系数中各对应变换系数的差值之和；根据所述第一目标值和所述第二目标值的加权系数，对所述第一目标值和所述第二目标值进行加权求和，得到所述图像差异值。
可选地，作为一个实施例，所述第二变换系数集是对所述第一图像和所述第二图像的差值图像进行所述变换得到的变换系数集，所述处理器420具体用于根据所述第一变换系数集与所述第二变换系数集中变换系数的一一对应关系，从所述第二变换系数集中选取与所述第一组变换系数对应的第二组变换系数；根据所述第二组变换系数，确定所述图像差异值。
可选地，作为一个实施例，所述处理器420具体用于将所述第二组变换系数中各变换系数之和确定为所述图像差异值。
可选地，作为一个实施例，所述第二组变换系数包括N个子组，所述处理器420具体用于确定N个目标值，其中，第i目标值为第i子组中的变换系数之和；根据所述N个目标值的加权系数，对所述N个目标值进行加权求和，得到所述图像差异值。
可选地，作为一个实施例，所述处理器420具体用于确定第一目标值，所述第一目标值为所述第二组变换系数中各变换系数之和；确定第二目标值，所述第二目标值为所述第二变换系数集中除所述第二组变换系数外的剩余 变换系数之和；根据所述第一目标值和所述第二目标值的加权系数，对所述第一目标值和所述第二目标值进行加权求和，得到所述图像差异值。
可选地，作为一个实施例，所述第一图像为目标图像，所述第二图像为K个候选图像中的任意候选图像，所述处理器420具体用于根据所述目标图像和每个候选图像的图像差异值，确定所述每个候选图像的加权系数；根据所述K个候选图像的像素值，以及所述K个候选图像的加权系数，确定所述目标图像的滤波图像。
可选地，作为一个实施例，所述处理器420具体用于根据
Figure PCTCN2016070228-appb-000016
确定所述每个候选图像的加权系数，其中，b1、a1和h1均表示正实数，Dk表示所述第一图像和第k个候选图像的图像差异值，Wk表示所述第k个候选图像的加权系数。
可选地，作为一个实施例，所述处理器420具体用于根据Wk＝b2-(Dk)a2/h2确定所述每个候选图像的加权系数，其中，b2、a2和h2均表示正实数，Dk表示所述第一图像和第k个候选图像的图像差异值，Wk表示所述第k个候选图像的加权系数。
可选地，作为一个实施例，所述处理器420具体用于根据
Figure PCTCN2016070228-appb-000017
确定所述每个候选图像的加权系数，其中，b3、a3和h3均表示正实数，Dk表示所述第一图像和第k个候选图像的图像差异值，Wk表示所述第k个候选图像的加权系数。
可选地，作为一个实施例，所述处理器420具体用于根据
Figure PCTCN2016070228-appb-000018
确定所述滤波图像，其中，Pfj表示所述滤波图像在第j像素点的像素值，W0表示所述第一图像的加权系数，Wk表示第k个候选图像的加权系数，P0j表示所述第一图像在第j像素点的像素值，Pkj表示所述第k个候选图像在第j像素点的像素值。
应理解，在本发明实施例中，术语“和/或”仅仅是一种描述关联对象的关联关系，表示可以存在三种关系。例如，A和/或B，可以表示：单独存在A，同时存在A和B，单独存在B这三种情况。另外，本文中字符“/”，一般表示前后关联对象是一种“或”的关系。
本领域普通技术人员可以意识到，结合本文中所公开的实施例描述的各示例的单元及算法步骤，能够以电子硬件、计算机软件或者二者的结合来实现，为了清楚地说明硬件和软件的可互换性，在上述说明中已经按照功能一般性地描述了各示例的组成及步骤。这些功能究竟以硬件还是软件方式来执行，取决于技术方案的特定应用和设计约束条件。专业技术人员可以对每个特定的应用来使用不同方法来实现所描述的功能，但是这种实现不应认为超出本发明的范围。
所属领域的技术人员可以清楚地了解到，为了描述的方便和简洁，上述描述的系统、装置和单元的具体工作过程，可以参考前述方法实施例中的对应过程，在此不再赘述。
在本申请所提供的几个实施例中，应该理解到，所揭露的系统、装置和方法，可以通过其它的方式实现。例如，以上所描述的装置实施例仅仅是示意性的，例如，所述单元的划分，仅仅为一种逻辑功能划分，实际实现时可以有另外的划分方式，例如多个单元或组件可以结合或者可以集成到另一个系统，或一些特征可以忽略，或不执行。另外，所显示或讨论的相互之间的耦合或直接耦合或通信连接可以是通过一些接口、装置或单元的间接耦合或通信连接，也可以是电的，机械的或其它的形式连接。
所述作为分离部件说明的单元可以是或者也可以不是物理上分开的，作为单元显示的部件可以是或者也可以不是物理单元，即可以位于一个地方，或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部单元来实现本发明实施例方案的目的。
另外，在本发明各个实施例中的各功能单元可以集成在一个处理单元中， 也可以是各个单元单独物理存在，也可以是两个或两个以上单元集成在一个单元中。上述集成的单元既可以采用硬件的形式实现，也可以采用软件功能单元的形式实现。
所述集成的单元如果以软件功能单元的形式实现并作为独立的产品销售或使用时，可以存储在一个计算机可读取存储介质中。基于这样的理解，本发明的技术方案本质上或者说对现有技术做出贡献的部分，或者该技术方案的全部或部分可以以软件产品的形式体现出来，该计算机软件产品存储在一个存储介质中，包括若干指令用以使得一台计算机设备(可以是个人计算机，服务器，或者网络设备等)执行本发明各个实施例所述方法的全部或部分步骤。而前述的存储介质包括：U盘、移动硬盘、只读存储器(ROM，Read-Only Memory)、随机存取存储器(RAM，Random Access Memory)、磁碟或者光盘等各种可以存储程序代码的介质。
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，可轻易想到各种等效的修改或替换，这些修改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以权利要求的保护范围为准。

Claims (28)

  1. 一种处理图像的方法，其特征在于，包括：
    对第一图像进行变换，得到第一变换系数集；
    对第二图像进行所述变换，或者对第一图像和第二图像的差值图像进行所述变换，得到第二变换系数集；
    根据所述第一变换系数集中的变换系数的幅度，从所述第一变换系数集中选取第一组变换系数，其中，所述第一组变换系数中的变换系数的幅度满足预设的阈值条件；
    根据所述第一组变换系数和所述第二变换系数集，确定第一图像和第二图像之间的图像差异值；
    根据所述图像差异值，对所述第一图像和所述第二图像进行处理。
  2. 如权利要求1所述的方法，其特征在于，所述第二变换系数集是对所述第二图像进行所述变换得到的变换系数集，
    所述根据所述第一组变换系数和所述第二变换系数集，确定第一图像和第二图像之间的图像差异值，包括：
    根据所述第一变换系数集与所述第二变换系数集中变换系数的一一对应关系，从所述第二变换系数集中选取与所述第一组变换系数对应的第二组变换系数；
    根据所述第一组变换系数和所述第二组变换系数，确定所述图像差异值。
  3. 如权利要求2所述的方法，其特征在于，所述根据所述第一组变换系数和所述第二组变换系数，确定所述图像差异值，包括：
    确定所述第一组变换系数和所述第二组变换系数中各对应变换系数的差值；
    将所述各对应变换系数的差值之和确定为所述图像差异值。
  4. 如权利要求2所述的方法，其特征在于，所述第一组变换系数和所述第二组变换系数均按照相同方式划分为N个子组，
    所述根据所述第一组变换系数和所述第二组变换系数，确定所述图像差异值，包括：
    确定N个目标值，其中，所述N个目标值中的第i目标值为所述第一组变换系数中的第i子组与所述第二组变换系数中的第i子组中各对应变换系数的差值之和；
    根据所述N个子组的加权系数，对所述N个目标值进行加权求和，得到所述图像差异值。
  5. 如权利要求2所述的方法，其特征在于，所述根据所述第一组变换系数和所述第二组变换系数，确定所述图像差异值，包括：
    确定第一目标值，其中，所述第一目标值为所述第一组变换系数与所述第二组变换系数中各对应变换系数的差值之和；
    确定第二目标值，其中，所述第二目标值为所述第一变换系数集中除所述第一组变换系数外的剩余变换系数与所述第二变换系数集中除所述第二组变换系数外的剩余变换系数中各对应变换系数的差值之和；
    根据所述第一目标值和所述第二目标值的加权系数，对所述第一目标值和所述第二目标值进行加权求和，得到所述图像差异值。
  6. 如权利要求1所述的方法，其特征在于，所述第二变换系数集是对所述第一图像和所述第二图像的差值图像进行所述变换得到的变换系数集，
    所述根据所述第一组变换系数和所述第二变换系数集，确定第一图像和第二图像之间的图像差异值，包括：
    根据所述第一变换系数集与所述第二变换系数集中变换系数的一一对应关系，从所述第二变换系数集中选取与所述第一组变换系数对应的第二组变换系数；
    根据所述第二组变换系数，确定所述图像差异值。
  7. 如权利要求6所述的方法，其特征在于，所述根据所述第二组变换系数，确定所述图像差异值，包括：
    将所述第二组变换系数中各变换系数之和确定为所述图像差异值。
  8. 如权利要求6所述的方法，其特征在于，所述第二组变换系数包括N个子组，
    所述根据所述第二组变换系数，确定所述图像差异值，包括：
    确定N个目标值，其中，第i目标值为第i子组中的变换系数之和；
    根据所述N个目标值的加权系数，对所述N个目标值进行加权求和，得到所述图像差异值。
  9. 如权利要求6所述的方法，其特征在于，所述根据所述第二组变换系数，确定所述图像差异值，包括：
    确定第一目标值，所述第一目标值为所述第二组变换系数中各变换系数之和；
    确定第二目标值，所述第二目标值为所述第二变换系数集中除所述第二组变换系数外的剩余变换系数之和；
    根据所述第一目标值和所述第二目标值的加权系数，对所述第一目标值和所述第二目标值进行加权求和，得到所述图像差异值。
  10. 如权利要求1所述的方法，其特征在于，所述第一图像为目标图像，所述第二图像为K个候选图像中的任意候选图像，
    所述根据所述图像差异值，对所述第一图像和所述第二图像进行处理，包括：
    根据所述目标图像和每个候选图像的图像差异值，确定所述每个候选图像的加权系数；
    根据所述K个候选图像的像素值，以及所述K个候选图像的加权系数，确定所述目标图像的滤波图像。
  11. 如权利要求10所述的方法，其特征在于，所述根据所述目标图像和每个候选图像的图像差异值，确定所述每个候选图像的加权系数，包括：
    根据
    Figure PCTCN2016070228-appb-100001
    确定所述每个候选图像的加权系数，其中，b1、a1和h1 均表示正实数，Dk表示所述第一图像和第k个候选图像的图像差异值，Wk表示所述第k个候选图像的加权系数。
  12. 如权利要求10所述的方法，其特征在于，所述根据所述目标图像和每个候选图像的图像差异值，确定所述每个候选图像的加权系数，包括：
    根据Wk＝b2-(Dk)a2/h2确定所述每个候选图像的加权系数，其中，b2、a2和h2均表示正实数，Dk表示所述第一图像和第k个候选图像的图像差异值，Wk表示所述第k个候选图像的加权系数。
  13. 如权利要求10所述的方法，其特征在于，所述根据所述目标图像和每个候选图像的图像差异值，确定所述每个候选图像对应的加权系数，包括：
    根据
    Figure PCTCN2016070228-appb-100002
    确定所述每个候选图像的加权系数，其中，b3、a3和h3均表示正实数，Dk表示所述第一图像和第k个候选图像的图像差异值，Wk表示所述第k个候选图像的加权系数。
  14. 如权利要求10所述的方法，其特征在于，所述根据所述K个候选图像的像素值，以及所述K个候选图像的加权系数，确定所述目标图像的滤波图像，包括：
    根据
    Figure PCTCN2016070228-appb-100003
    确定所述滤波图像，其中，Pfj表示所述滤波图像在第j像素点的像素值，W0表示所述第一图像的加权系数，Wk表示第k个候选图像的加权系数，P0j表示所述第一图像在第j像素点的像素值，Pkj表示所述第k个候选图像在第j像素点的像素值。
  15. 一种处理图像的装置，其特征在于，包括：
    第一变换单元，用于对第一图像进行变换，得到第一变换系数集；
    第二变换单元，用于对第二图像进行所述变换，或者对第一图像和第二图像的差值图像进行所述变换，得到第二变换系数集；
    选取单元，用于根据所述第一变换单元得到的所述第一变换系数集中的变换系数的幅度，从所述第一变换系数集中选取第一组变换系数，其中，所述第一组变换系数中的变换系数的幅度满足预设的阈值条件；
    确定单元，用于根据所述选取单元选取的所述第一组变换系数和所述第二变换单元得到的所述第二变换系数集，确定第一图像和第二图像之间的图像差异值；
    处理单元，用于根据所述确定单元确定的所述图像差异值，对所述第一图像和所述第二图像进行处理。
  16. 如权利要求15所述的装置，其特征在于，所述第二变换系数集是对所述第二图像进行所述变换得到的变换系数集，所述确定单元具体用于根据所述第一变换系数集与所述第二变换系数集中变换系数的一一对应关系，从所述第二变换系数集中选取与所述第一组变换系数对应的第二组变换系数；根据所述第一组变换系数和所述第二组变换系数，确定所述图像差异值。
  17. 如权利要求16所述的装置，其特征在于，所述确定单元具体用于确定所述第一组变换系数和所述第二组变换系数中各对应变换系数的差值；将所述各对应变换系数的差值之和确定为所述图像差异值。
  18. 如权利要求16所述的装置，其特征在于，所述第一组变换系数和所述第二组变换系数均按照相同方式划分为N个子组，所述确定单元具体用于确定N个目标值，其中，所述N个目标值中的第i目标值为所述第一组变换系数中的第i子组与所述第二组变换系数中的第i子组中各对应变换系数的差值之和；根据所述N个子组的加权系数，对所述N个目标值进行加权求和，得到所述图像差异值。
  19. 如权利要求16所述的装置，其特征在于，所述确定单元具体用于确定第一目标值，其中，所述第一目标值为所述第一组变换系数与所述第二组变换系数中各对应变换系数的差值之和；确定第二目标值，其中，所述第二目标值为所述第一变换系数集中除所述第一组变换系数外的剩余变换系 数与所述第二变换系数集中除所述第二组变换系数外的剩余变换系数中各对应变换系数的差值之和；根据所述第一目标值和所述第二目标值的加权系数，对所述第一目标值和所述第二目标值进行加权求和，得到所述图像差异值。
  20. 如权利要求15所述的装置，其特征在于，所述第二变换系数集是对所述第一图像和所述第二图像的差值图像进行所述变换得到的变换系数集，所述确定单元具体用于根据所述第一变换系数集与所述第二变换系数集中变换系数的一一对应关系，从所述第二变换系数集中选取与所述第一组变换系数对应的第二组变换系数；根据所述第二组变换系数，确定所述图像差异值。
  21. 如权利要求20所述的装置，其特征在于，所述确定单元具体用于将所述第二组变换系数中各变换系数之和确定为所述图像差异值。
  22. 如权利要求20所述的装置，其特征在于，所述第二组变换系数包括N个子组，所述确定单元具体用于确定N个目标值，其中，第i目标值为第i子组中的变换系数之和；根据所述N个目标值的加权系数，对所述N个目标值进行加权求和，得到所述图像差异值。
  23. 如权利要求20所述的装置，其特征在于，所述确定单元具体用于确定第一目标值，所述第一目标值为所述第二组变换系数中各变换系数之和；确定第二目标值，所述第二目标值为所述第二变换系数集中除所述第二组变换系数外的剩余变换系数之和；根据所述第一目标值和所述第二目标值的加权系数，对所述第一目标值和所述第二目标值进行加权求和，得到所述图像差异值。
  24. 如权利要求15所述的装置，其特征在于，所述第一图像为目标图像，所述第二图像为K个候选图像中的任意候选图像，所述处理单元具体用于根据所述目标图像和每个候选图像的图像差异值，确定所述每个候选图像的加权系数；根据所述K个候选图像的像素值，以及所述K个候选图像的 加权系数，确定所述目标图像的滤波图像。
  25. 如权利要求24所述的装置，其特征在于，所述确定单元具体用于根据
    Figure PCTCN2016070228-appb-100004
    确定所述每个候选图像的加权系数，其中，b1、a1和h1均表示正实数，Dk表示所述第一图像和第k个候选图像的图像差异值，Wk表示所述第k个候选图像的加权系数。
  26. 如权利要求24所述的装置，其特征在于，所述确定单元具体用于根据Wk＝b2-(Dk)a2/h2确定所述每个候选图像的加权系数，其中，b2、a2和h2均表示正实数，Dk表示所述第一图像和第k个候选图像的图像差异值，Wk表示所述第k个候选图像的加权系数。
  27. 如权利要求24所述的装置，其特征在于，所述确定单元具体用于根据
    Figure PCTCN2016070228-appb-100005
    确定所述每个候选图像的加权系数，其中，b3、a3和h3均表示正实数，Dk表示所述第一图像和第k个候选图像的图像差异值，Wk表示所述第k个候选图像的加权系数。
  28. 如权利要求24所述的装置，其特征在于，所述处理单元具体用于根据
    Figure PCTCN2016070228-appb-100006
    确定所述滤波图像，其中，Pfj表示所述滤波图像在第j像素点的像素值，W0表示所述第一图像的加权系数，Wk表示第k个候选图像的加权系数，P0j表示所述第一图像在第j像素点的像素值，Pkj表示所述第k个候选图像在第j像素点的像素值。
PCT/CN2016/070228 2015-01-09 2016-01-06 处理图像的方法和装置 WO2016110249A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
EP16734909.1A EP3236658B1 (en) 2015-01-09 2016-01-06 Image processing method and device determining an image difference using transform coefficients
US15/645,052 US10291935B2 (en) 2015-01-09 2017-07-10 Image processing method and apparatus

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201510012142.4A CN104602025B (zh) 2015-01-09 2015-01-09 处理图像的方法和装置
CN201510012142.4 2015-01-09

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US15/645,052 Continuation US10291935B2 (en) 2015-01-09 2017-07-10 Image processing method and apparatus

Publications (1)

Publication Number Publication Date
WO2016110249A1 true WO2016110249A1 (zh) 2016-07-14

Family

ID=53127461

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2016/070228 WO2016110249A1 (zh) 2015-01-09 2016-01-06 处理图像的方法和装置

Country Status (4)

Country Link
US (1) US10291935B2 (zh)
EP (1) EP3236658B1 (zh)
CN (1) CN104602025B (zh)
WO (1) WO2016110249A1 (zh)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104602025B (zh) 2015-01-09 2018-11-20 华为技术有限公司 处理图像的方法和装置
US10869060B2 (en) * 2018-01-30 2020-12-15 Google Llc Efficient context model computation design in transform coefficient coding
CN109242011A (zh) * 2018-08-27 2019-01-18 深圳开立生物医疗科技股份有限公司 一种识别图像差异的方法及装置
CN111416977B (zh) * 2019-01-07 2024-02-09 浙江大学 视频编码器、视频解码器及相应方法

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5396292A (en) * 1992-02-26 1995-03-07 Nec Corporation Encoding apparatus for motion video signals
CN1145564A (zh) * 1995-09-12 1997-03-19 大宇电子株式会社 用于编码目标轮廓的装置
CN101409845A (zh) * 2008-10-31 2009-04-15 北京大学软件与微电子学院 一种avs视频编码中的视频失真度估计方法及其装置
CN103442235A (zh) * 2013-09-06 2013-12-11 深圳市融创天下科技股份有限公司 一种图像处理方法以及装置
US20140098861A1 (en) * 2012-10-05 2014-04-10 Futurewei Technologies, Inc. Architecture for Hybrid Video Codec
CN104602025A (zh) * 2015-01-09 2015-05-06 华为技术有限公司 处理图像的方法和装置

Family Cites Families (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20040086152A1 (en) * 2002-10-30 2004-05-06 Ramakrishna Kakarala Event detection for video surveillance systems using transform coefficients of compressed images
US20060285590A1 (en) * 2005-06-21 2006-12-21 Docomo Communications Laboratories Usa, Inc. Nonlinear, prediction filter for hybrid video compression
JP4555257B2 (ja) * 2006-06-06 2010-09-29 パナソニック株式会社 画像符号化装置
US8396325B1 (en) * 2009-04-27 2013-03-12 Google Inc. Image enhancement through discrete patch optimization
KR101545382B1 (ko) * 2010-01-08 2015-08-18 노키아 코포레이션 비디오 코딩을 위한 장치, 방법 및 컴퓨터 프로그램
CN101783954B (zh) * 2010-03-12 2012-08-08 厦门大学 一种视频图像编解码方法
US9451271B2 (en) * 2011-07-21 2016-09-20 Blackberry Limited Adaptive filtering based on pattern information
CN103108177B (zh) * 2011-11-09 2016-11-23 华为技术有限公司 图像编码方法及图像编码装置
EP2823460B1 (en) * 2012-03-05 2018-05-23 Thomson Licensing DTV Method and apparatus for performing hierarchical super-resolution of an input image
CN104350746A (zh) * 2012-05-31 2015-02-11 汤姆逊许可公司 基于局部幅度和相位谱的图像质量测量
US9344742B2 (en) * 2012-08-10 2016-05-17 Google Inc. Transform-domain intra prediction
US9171226B2 (en) * 2012-09-26 2015-10-27 Carnegie Mellon University Image matching using subspace-based discrete transform encoded local binary patterns
CN103974076B (zh) * 2014-05-19 2018-01-12 华为技术有限公司 图像编解码方法和设备、系统

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5396292A (en) * 1992-02-26 1995-03-07 Nec Corporation Encoding apparatus for motion video signals
CN1145564A (zh) * 1995-09-12 1997-03-19 大宇电子株式会社 用于编码目标轮廓的装置
CN101409845A (zh) * 2008-10-31 2009-04-15 北京大学软件与微电子学院 一种avs视频编码中的视频失真度估计方法及其装置
US20140098861A1 (en) * 2012-10-05 2014-04-10 Futurewei Technologies, Inc. Architecture for Hybrid Video Codec
CN103442235A (zh) * 2013-09-06 2013-12-11 深圳市融创天下科技股份有限公司 一种图像处理方法以及装置
CN104602025A (zh) * 2015-01-09 2015-05-06 华为技术有限公司 处理图像的方法和装置

Also Published As

Publication number Publication date
CN104602025A (zh) 2015-05-06
US20170310998A1 (en) 2017-10-26
EP3236658A1 (en) 2017-10-25
CN104602025B (zh) 2018-11-20
EP3236658B1 (en) 2019-08-14
EP3236658A4 (en) 2018-01-03
US10291935B2 (en) 2019-05-14

Similar Documents

Publication Publication Date Title
US11381844B2 (en) Method and apparatus for encoding/decoding images using a prediction method adopting in-loop filtering
KR101874102B1 (ko) 영상의 인트라 예측 부호화 방법 및 장치 및 컴퓨터 기록매체
CN111819852B (zh) 用于变换域中残差符号预测的方法及装置
JP5128443B2 (ja) 映像の符号化、復号化の方法及びその装置
CN105284112B (zh) 用于确定量化参数的值的方法和装置
CN107046645B (zh) 图像编解码方法及装置
CN113228646A (zh) 具有非线性限幅的自适应环路滤波(alf)
TWI489878B (zh) 影像編碼方法、影像解碼方法、影像編碼裝置、影像解碼裝置及該等之程式
WO2016110249A1 (zh) 处理图像的方法和装置
Chao et al. Edge-adaptive depth map coding with lifting transform on graphs
CN110741642B (zh) 使用拟合平面和参考样本进行定向帧内预测的装置和方法
US20200267409A1 (en) Early termination of block-matching for collaborative filtering
Tabus et al. Sparse prediction for compression of stereo color images conditional on constant disparity patches
Roy et al. Graph-based transform with weighted self-loops for predictive transform coding based on template matching
Ding et al. Context-based adaptive zigzag scanning for image coding
Lee et al. A new intra prediction method using channel correlations for the H. 264/AVC intra coding
Wang et al. Graph Based Cross-Channel Transform for Color Image Compression
Sun et al. An efficient DCT-based image compression system based on transparent composite model
CN110710209A (zh) 运动补偿的方法、装置和计算机系统
Huang et al. A weighted low-rank matrix approximation based template matching scheme for inter-frame prediction
Wu et al. An improved intra-prediction algorithm for H. 264

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: 16734909

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

REEP Request for entry into the european phase

Ref document number: 2016734909

Country of ref document: EP