FIG. 7 is a flowchart illustrating an exemplary method for lossless ARGB compression based on an intra-block prediction according to some embodiments of the present disclosure.

In some embodiments, the processor may be configured to perform process 700, as shown in FIG. 7. The process 700 is an exemplary process of the method for lossless ARGB compression based on the intra-block prediction. The process 700 includes steps 710-740, and the processor may be configured to perform the steps 710-740 for each input image block under each processing channel until all input image blocks are complete encoded and perform step 740 based on the results of the above steps. Steps 710-740 are shown below.

In 710, a predicted value of the input image block is obtained under a current processing channel based on the input image block under the current processing channel by the intra-block prediction.

The input image block refers to an image that includes a part of the input image. The input image refers to a target image to be compressed.

In some embodiments, the processor may determine at least one input image block based on the input image.

For example, the processor may divide the input image according to a preset shape to determine at least one input image block. The preset shape may be a regular shape or an irregular shape. The preset shape may be set according to the demand. For example, the preset shape may be a rectangle of n×m, or the like. Where n and m denote pixel values, and values of n and m may be equal or unequal.

As another example, the processor may divide the input image by a preset number. The preset number is related to image complexity. The more complex the input image, the larger the preset number. The image complexity is used to indicate the complexity of the input image. The image complexity may be expressed as the number of colors in the input image. The more colors in the input image, the higher the image complexity. The number of colors may be obtained from a color histogram, a number of discrete colors, or other color quantization manners.

In some embodiments, the input image may be obtained based on a user terminal. The user terminal refers to a terminal server used by a user, e.g., a mobile phone, a personal computer, a tablet, or the like. The processor may obtain the input image uploaded by the user based on the user terminal.

In some embodiments, the input image block is determined based on the input image, the input image is an ARGB image. In some embodiments, the input image block is obtained by dividing the ARGB image into 8*8 sized image blocks. Where 8 indicates that the number of pixels is 8.

The ARGB image is an image format that includes ARGB components. The ARGB components include a transparency, an intensity of a red component, an intensity of a green component, and an intensity of a blue component.

In some embodiments, the processor may divide the ARGB image into 8*8 sized image blocks to obtain at least one input image block, and each input image block is the ARGB image.

In some embodiments, the processor may also determine at least one input image block by determining whether a first edge feature or a second edge feature exists in the input image. More detailed descriptions may be found in related descriptions hereinafter.

The processing channel refers to a channel that process different parameters of an image. In some embodiments, the processing channel may be ARGB channels, including an A channel, an R channel, a G channel, and a B channel. More detailed descriptions regarding the ARGB channels may be found in related descriptions hereinabove. The current processing channel refers to one of the processing channels currently used to process the image.

The intra-block prediction is a manner that performs pixel prediction separately for each input image block. More detailed descriptions regarding the intra-block prediction may be found in the related descriptions hereinabove.

In some embodiments, the intra-block prediction includes a prediction by dividing the input image block into three parts.

In some embodiments, the processor may randomly divide the input image block into three parts for prediction. In some embodiments, the processor may divide the input image block into three parts for prediction according to a preset size. The preset size may be predetermined according to demand.

Exemplarily, as shown in FIG. 4, a first part 10 of the input image block includes original pixels in the upper left corner, a second part 20 includes pixels in a first row and a first column excluding the original pixels in the upper left corner, and a third part 30 includes the remaining pixels in the input image block excluding the first part and the second part.

The accuracy of the pixel prediction of the input image block may be improved by further dividing the input image block.

In some embodiments, the input image block includes at least one row pixel group and at least one column pixel group. The processor may determine a similarity set and divide the input image block into three parts based on the similarity set.

The row pixel group refers to a collection of pixel values for a row in the input image block. The row pixel group may be represented as a vector ensemble. Exemplarily, the row pixel group may be represented as (a₁₁, a₁₂, a₁₃. . . a_1y), (a₂₁, a₂₂, a₂₃. . . a_2y), . . . , (a_x1, a_x2, a_x3. . . a_xy), where, a_xyindicates the pixel value in row x and column y.

The column pixel group refers to a collection of pixel values of a column in the input image block. The column pixel group may be represented as a vector. Exemplarily, the total number of rows of pixels in the input image block is x, the total number of columns is y, and the column pixel group may be represented as (a₁₁, a₂₁, a₃₁. . . a_x1), (a₁₂, a₂₂, a₃₂. . . a_x2), . . . , (a_1y, a_2y, a_3y. . . a_xy).

The similarity set refers to an ensemble representing the similarity of rows and columns of the input image block. In some embodiments, the similarity set includes similarities of the at least one row pixel group and the at least one column pixel group of the input image block. The similarity set also includes a label corresponding to the similarity. The similarity set may be represented as a vector ensemble. Exemplarily, the similarity set may be represented as [(0.4, 1, 2), (0.3, 1, 3), or the like]. The similarity is 0.4 in (0.4, 1, 2), the label is (1, 2), and (0.4, 1, 2) indicates that the similarity between the first row and the second column is 0.4. The similarity in (0.3, 1, 3) is 0.3, the label is (1, 3), and (0.3, 1, 3) indicates that the similarity between the first row and the third column is 0.3.

In some embodiments, the processor may compute a similarity between the row pixel group and the column pixel group, and take the similarity, a row label of the row pixel group, and a column label of the column pixel group as the similarity set. The similarity is negatively correlated with a vector distance between a row vector and a column vector. The row vector refers to a vector corresponding to a row in the row pixel group. The column vector refers to a vector corresponding to a column in the column pixel group. The vector distance may be expressed as a cosine distance or a Euclidean distance.

In some embodiments, the processor may group the input image block based on the similarity set through a variety of manners.

For example, the processor may select a row and column of the input image block that have the highest similarity in the similarity set, take intersection pixels as a first part, take the remaining pixels in the same row and column as a second part, and take pixels in the input image block excluding the first part and the second part as a third part.

Exemplarily, the row and column of the highest similarity in the similarity set is (0.4, 1, 2), that means the similarity between the first row and the second column is the highest. The intersection pixel of the first row and the second column is a₁₂, then a₁₂is used as the first part, and pixels in the first row and the second column excluding a₁₂are used as the second part, and the other pixels excluding the pixels in the first part and the second part are used as the third part.

Grouping by the similarity between rows and columns of the input image block may make the obtained group more reasonable, which makes the subsequent prediction more accurate.

The predicted value of the input image block refers to a pixel value of the input image obtained by prediction.

In some embodiments, the processor may predict, based on the input image block under the current processing channel, to obtain the predicted value of the input image block under the current processing channel through various manners.

For example, for each input image block under the current processing channel, the processor may separately predict to obtain the predicted value of the input image block under the current processing channel by a preset algorithm. The preset algorithm may be at least one of a differential pulse code modulation (DPCM) algorithm, a median edge detector (MED) algorithm, other pixel prediction algorithms, or the like.

In some embodiments, the processor may predict each of the three parts of the input image block, and the pixel values obtained from the prediction of the three parts of the input image block are used as the predicted values of the input image block. For each part of the input image block, at least one prediction algorithm is used, or the original pixel value is retained. The prediction algorithms may be the same or different for each part.

Exemplarily, as shown in FIG. 4, an input image block includes the first part 10, the second part 20, and the third part 30. The processor may retain the original pixel value of the first part 10. The processor may apply a DPCM algorithm to the second part 20 for prediction and designate a plurality of predicted pixel values as first predicted values. The processor may apply the MED algorithm to the third part 30 for prediction and designate a plurality of pixel values as second predicted values. The processor may output the original pixel value of the first part, the first predicted values of the second part, and the second predicted values of the third part as the predicted values of the input image block. More detailed descriptions may be found in FIG. 4 and the related descriptions.

In some embodiments, the processor may perform a differential pulse code modulation (DPCM) prediction on a first row and a first column of the input image block and perform a median edge detector (MED) prediction on a region of the input image block after removing the first row and the first column. More detailed descriptions may be found in FIG. 5 and related descriptions thereof.

In some embodiments, the processor may use a plurality of pixel values obtained by prediction using the DPCM algorithm and the MED algorithm as the predicted values of the input image block.

The DPCM algorithm is simple but sensitive to noise, which may cause error accumulation. The MED algorithm has high accuracy and good noise suppression but high computational complexity. Using a combination of the DPCM algorithm and the MED algorithm, the compression effect of the image may be better by utilizing the redundant information of the image in two directions.

In some embodiments, the processor may also determine a second pixel value of the input image block based on a first pixel value sequence of the input image block. The first pixel value sequence refers to a sequence consisting of actual pixel values that lie within a predetermined range of the second pixel value. The second pixel value is a predicted value of a pixel of the input image block currently predicted. More detailed descriptions may be found in FIG. 8 and related descriptions thereof.

In 720, predicted residuals of the input image block are determined under the current processing channel based on the predicted value and the original pixel value.

The original pixel value refers to a true pixel value of the input image block obtained by measuring. In some embodiments, the processor may obtain an image uploaded by a user based on a user terminal that contains the original pixel value and extract the original pixel value in the image. The user terminal may be a terminal device such as a mobile phone, a personal computer, a tablet computer, or the like.

The predicted residuals are used to indicate a difference between the predicted value of a pixel and the original pixel value of the input image block.

In some embodiments, the processor may determine, based on the predicted value and the original pixel value, the predicted residuals of the input image block under the current processing channel through a plurality of manners.

For example, the processor may use the absolute value of the difference between the original pixel value and the predicted value as the predicted residual.

In 730, the predicted residuals are input into a residual encoder for encoding to obtain a residual stream of the input image block under the current processing channel.

The residual encoder refers to a processor including a predetermined encoding program. The residual encoder may be configured as at least one of a digital signal processor (DSP), an image processing unit (IPU), a central processing unit (CPU), or the like. The predetermined encoding program refers to a predetermined program for encoding the predicted residuals. For example, the predetermined encoding program may be a program that encodes the predicted residuals using small value optimized variable length encoding.

The residual stream refers to a bit stream formed by encoding the predicted residuals of the input image block. By encoding the predicted residuals into a code stream, the amount of data may be effectively reduced, thereby saving storage space and transmission bandwidth. When decoding, the original image may be recovered by decoding the residual stream and combining the predicted value.

In some embodiments, the processor may input the predicted residuals into the residual encoder, which encodes them through a variety of manners to obtain the residual stream of the input image block under the current processing channel. For example, the processor may use the small value optimized variable length encoding to encode the predicted residuals.

In some embodiments, the processor may also encode the predicted residuals based on other manners, such as Huffman coding, arithmetic coding, or the like.

In some embodiments, the processor may group the predicted residuals of the input image block. For example, the processor may randomly group the predicted residuals. A count of groups may be set as desired. As another example, the processor may take the predicted residuals corresponding to each row of pixels of the input image block under the current processing channel as a residual group. As yet another example, the processor may take the predicted residuals corresponding to each column of pixels of the input image block under the current processing channel as a residual group. The residual group refers to an ensemble consisting of a plurality of predicted residuals.

Grouping the predicted residuals may increase the accuracy of subsequent encoding.

In some embodiments, the processor may obtain at least one residual group by grouping the predicted residuals based on a predetermined condition.

The predetermined condition may be set based on demand or experience. The predetermined condition may include that predicted residuals in the same column or row that are less than a preset threshold are used as a residual group and predicted residuals in the same column or row that are greater than or equal to the preset threshold are used as a residual group. Exemplarily, the predetermined condition is that the distribution is similar in the horizontal direction, i.e., the predicted residuals in the horizontal direction are less than the preset threshold. The preset threshold may be set according to experience.

In some embodiments, for the same residual group, the processor may determine a header of the residual group based on a maximum absolute value of the residual group.

The encoding unit refers to the smallest unit used for encoding. The residual group refers to a collection of at least one predicted residual corresponding to the same input image block. The header is a symbol or code used to determine an encoding group of the residual group in an encoding table. The encoding table is a table including an encoding rule. Exemplarily, the encoding table may be an SVO-VLC code table. The same predicted residual is encoded differently under different headers. In some embodiments, the processor may divide the maximum absolute value into different predetermined ranges, with each predetermined range corresponding to a header. Exemplarily, as shown in the SVO-VLC code table, the predetermined range may include 0, 1, 2-3, 4-7, 8-15, 16-31, or greater than 31, with corresponding headers including 01, 10, 110, 1110, 11110, 111110, or 111111. Header 110 represents that the maximum absolute value of the encoding unit is within a range of 2-3. If the maximum absolute value of predicted residuals in a residual group is 2 or 3, all predicted residuals in the residual group correspond to codes under a column with header 110. More detailed descriptions regarding SVO-VLC code table may be found in related descriptions hereinabove.

The encoding rule of the predicted residuals may be preset, with each predicted residual corresponding to a unique code. Exemplarily, the encoding rule of the predicted residuals may refer to the SVO-VLC code table. In the last bit of the code, 0 indicates that the predicted residual is positive and 1 indicates that the predicted residual is negative. For example, when a residual group is (1, −2), the maximum absolute value of the predicted residuals is 2, then code in the “2-3” column is used. A predicted residual 1 corresponds to code 010 and a predicted residual 2 corresponds to code 101, the residual group corresponds to code (010, 101).

In some embodiments, the processor may, in response to determining that the predicted residuals of the residual group are all first predetermined values, output the code of the header; or in response to determining that a maximum value of the predicted residuals of the residual group is greater than a second predetermined value, directly output an original pixel value corresponding to the residual group.

In some embodiments, the first predetermined value is a set value of the predicted residuals. Exemplarily, as shown in the SVO-VLC code table, the first predetermined value may be set to 0, i.e., when the predicted residuals of a residual group are all 0, the code 01 of the header is output.

The second predetermined value refers to a threshold of the maximum absolute value of the predicted residuals. Exemplarily, as shown in the SVO-VLC code table, the second predetermined value may be 31, i.e., when the maximum absolute value of the predicted residuals of a residual group is greater than 31, the original pixel value corresponding to the residual group is directly output.

In 740, residual streams of all input image blocks of all processing channels are stored in a storage file as a compressed file.

In some embodiments, the compressed file may be in a ZIP format, a RAR format, or the like. The compressed file may achieve full data recovery based on a residual encoding stream, indicating that the compression of the input image block is lossless.

A special encoding is performed by setting the first predetermined value and the second predetermined value to simplify the input codes and improve compression efficiency when all the predicted residuals of the residual group are 0. When the predicted residuals of the residuals group are relatively large as a whole, it indicates that there is a significant error in the predicted pixel value, then the original pixel value is output to ensure that the image is accurate and is not distorted.

In some embodiments, determining at least one input image block based on the input image includes determining, based on the input image, whether a first edge feature exists; in response to the existence of the first edge feature, determining the at least one input image block based on the first edge feature; in response to the absence of the first edge feature, determining a first image feature based on the input image, and determining the at least one input image block based on the first image feature.

The first edge feature refers to an edge feature of the full channel image of the input image. The edge feature is a region of the image that is significantly varied and represents a significant change in the boundary, structure, or texture of an object in the input image. For example, the first edge feature includes discontinuities in the distribution of an image characteristic (e.g., a pixel grayscale, a color, a texture, etc.) in the input image and/or a collection of pixels with a step change or a ridge-like change in the image characteristic.

In some embodiments, the processor may determine whether the first edge feature exists based on the input image by an edge detection algorithm. The edge detection algorithm may be at least one of a Canny operator, a LOG operator, a Roberts edge detection operator, or other edge detection algorithms.

In some embodiments, the processor may determine at least one input image block based on the first edge feature in response to the existence of the first edge feature through a variety of manners. For example, the processor may designate a consecutive part of the first edge feature as a closed edge, with pixels contained within each closed edge divided into one input image block.

The first image feature refers to a feature of each pixel in the full channel image. For example, a brightness, a color value, etc., of the pixel.

In some embodiments, the processor may, in response to the absence of the first edge feature, determine the first image feature based on the input image and determine the at least one input image block based on the first image feature through a variety of manners.

For example, the processor may extract the brightness, the color value, or the like of each pixel in the input image as the first image feature of the input image. The processor may designate a pixel having a similarity greater than a preset threshold and a consecutive plurality of pixels as an input image block based on a similarity of first image features of two adjacent pixels in the input image. The processor may construct a feature vector based on at least one of the brightness and the color value of each input image, and determine the similarity based on a vector distance between the feature vectors of the two adjacent pixels. The similarity is negatively correlated with the vector distance. The preset threshold may be set empirically. For example, the processor may use a statistical value of the similarity corresponding to all pairs of adjacent pixels in the input image as the preset threshold. Exemplarily, the statistical value may be an average value, or the like.

Pixel values in an image typically have strong correlation over a small local range, i.e., spatially neighboring pixels tend to have similar color or brightness. Chunking based on the first edge feature and the first image feature may more efficiently utilize the local correlation for more efficient compression of the input image.

It should be noted that the foregoing description of the process 700 is intended to be exemplary and illustrative only and does not limit the scope of application of the present disclosure. For those skilled in the art, various corrections and changes may be made to the process 700 under the guidance of the present disclosure. However, these corrections and changes remain within the scope of the present disclosure.

In some embodiments, determining the at least one input image block based on the input image includes determining, for each processing channel, based on the input image and the processing channel, whether a second edge feature exists; in response to the existence of the second edge feature, determining, based on the second edge feature, at least one input image block; in response to the absence of the second edge feature, determining, based on the input image and the processing channel, the second image feature pf the processing channel; and determining the at least one input image block based on second image features of all processing channels.

The second edge feature refers to an edge feature of a single-channel image of the input image. The single-channel image refers to an input image with only one processing channel turned on. In some embodiments, the processor may sequentially enable one of the processing channels for the input image and block the other processing channels to obtain the single-channel image of the input image. Descriptions regarding the edge features may be found in related descriptions hereinabove.

In some embodiments, for each processing channel, the processor may, based on the single-channel image of the input image, determine whether the second edge feature exists by the edge detection algorithm. In response to the existence of the second edge feature, the processor may designate a consecutive part in the second edge feature as a closed edge, where pixels contained within each closed edge are divided into the input image block. The manner for determining the second edge feature is similar to the manner for determining the first edge feature, more detailed descriptions may be found in related descriptions hereinabove.

The second image feature refers to a feature of each pixel in the single-channel image. For example, a brightness, a color value, or the like, of a pixel.

In some embodiments, the processor may determine the second image feature based on the input image in response to the absence of the second edge feature and determine the at least one input image block based on the second image feature through a variety of manners.

For example, for each processing channel, the processor may extract the brightness, the color value, or the like, of each pixel in the input image under the processing channel as the second image feature of the input image. The processor may designate a plurality of pixels whose similarity is greater than a preset threshold and are consecutive as an input image block under the processing channel based on the similarity of the first image feature of two adjacent pixels in the input image. The specific manner is similar to the manner for determining the input image block based on the first image feature, and more detailed descriptions may be found in the related descriptions hereinabove.

Each processing channel is independently predicted and encoded. Image cutting based on the features of the single-channel image of the input image may more effectively utilize the local correlation of the single-channel image of the input image and may perform efficient compression for each processing channel.

In some embodiments, obtaining the predicted value of the input image block under the current processing channel based on the input image block under the current processing channel by the intra-block prediction includes determining, based on a first pixel value sequence of the input image block, a second pixel value of the input image block.

The second pixel value refers to a predicted value of a pixel of the input image block currently predicted.

The first pixel value refers to an original pixel value of a pixel located within a predetermined range of the second pixel value. The first pixel value sequence refers to a sequence consisting of actual pixel values located within the predetermined range of the second pixel value. The first pixel value sequence does not include the original pixel value of the pixel where the second pixel value is located. More detailed descriptions regarding the original pixel value may be found in FIG. 7 and related descriptions thereof.

The predetermined range maybe preset according to the demand or experience. Using a smaller predetermined range may not be able to represent the features of the input image block, and adopting a larger predetermined range may increase the complexity of the calculation. Based on the principle of guaranteeing the features of the input image block, a smaller range should be used as much as possible. If the second pixel value is located at the edge of the input image block and the predetermined range is beyond the input image block, the excess may be supplemented by 0.

Exemplarily, the predetermined range may be 3*3, and the processor may take a sequence of actual pixel values excluding the second pixel value within the 3*3 range, centered on the pixel corresponding to the second pixel value, as the first pixel value sequence.

In some embodiments, the processor may determine the second pixel value of the input image block based on the first pixel value sequence of the input image block through a variety of manners. For example, the processor may use a statistic value of the first pixel value sequence as the second pixel value. Exemplarily, the statistic value may be an average, or the like.

The first pixel value sequence contains pixel information near the second pixel value. Determining the second pixel value by the first pixel value sequence may fully utilize the information of the surrounding pixel values to make the predicted value of the input image block more accurate.

FIG. 8 is a schematic diagram illustrating an exemplary prediction model according to some embodiments of the present disclosure.