Context-based reshaping algorithms for encoding video data
Systems and methods for context-based encoding of video data using reshaping algorithms. One method includes receiving the video data, the video data composed of a plurality of image frames, each image frame including a plurality of pixel blocks. The method includes determining, for each pixel block, a luma bin index, determining, for each luma bin, a banding risk value, and determining Gaussian function parameters based on the banding risk value. The method includes generating a differential reshaping function using the Gaussian function parameters, computing a luma-based forward reshaping function based on the differential reshaping function, and generating an output image for each image frame by applying the luma-based forward reshaping function to the respective image frame.
1. A video delivery system for encoding of HDR video data, the delivery system comprising:
a processor to perform encoding of HDR video data, the processor configured to:
receive the HDR video data, the HDR video data composed of a plurality of image frames, each image frame including a plurality of pixel blocks;
compute, for each pixel block k, a mean for luma pixel values included in the pixel block k;
compute, for each pixel block k, a standard deviation for luma pixel values included in the pixel block k;
determine, for each pixel block k, a block-mean bin index b k , wherein the entire HDR luma codeword range is divided into N B non-overlapping luma bins, wherein each luma bin b represents a corresponding discrete luma codeword range, and wherein the block-mean bin index b k is the floor function of multiplying the mean for luma pixel values included in the pixel block k with N B ;
compute, for each luma bin b, a block histogram by counting a number of pixels that have a block-mean bin index b k equal to b;
compute, for each luma bin b, a block standard deviation by averaging standard deviation over all pixel blocks that have a block-mean bin index b k equal to b;
determine, for each luma bin b, a banding risk value, wherein the banding risk value is determined using a multiplicative combination of the block histogram and the block standard deviation of the respective luma bin;
determine a Gaussian function defined by Gaussian function parameters including a mean value and a width value of a corresponding Gaussian distribution, wherein the width value is the reciprocal of the doubled square of the standard deviation of the Gaussian distribution, wherein the mean value of the Gaussian function parameters is set to the maximum banding risk value across all luma bins;
generate a differential reshaping function using the Gaussian function parameters, the differential reshaping function specifying the amount of increment to the next lower input value of the differential reshaping function;
compute a luma-based forward reshaping function based on the differential reshaping function, wherein the luma-based forward reshaping function is a monotonically non-decreasing function that transfers a higher bit-depth codeword to a lower bit-depth by using the differential reshaping function; and
generate an output image for each image frame by applying the luma-based forward reshaping function to the respective image frame.
2. The video delivery system according to claim 1 , wherein the width value of the Gaussian function parameters is selected based on image statistics such that the most banding prone luminance range is covered by the Gaussian function.
3. The video delivery system according to claim 1 , wherein the processor is further configured to:
store a normalized base-layer codeword as a single-channel forward reshaping function.
4. The video delivery system according to claim 1 , wherein the processor is further configured to:
determine a backwards reshaping function based on the luma-based forward reshaping function.
5. The video delivery system according to claim 4 , wherein the backwards reshaping function is approximated in the form of an 8-piece 1 st order polynomial curve.
6. The video delivery system according to claim 1 , wherein the differential reshaping function defines a number of codewords allocated to a given luminance range.
7. A method for encoding of HDR video data, the method comprising:
receiving the HDR video data, the HDR video data composed of a plurality of image frames, each image frame including a plurality of pixel blocks;
computing, for each pixel block k, a mean for luma pixel values included in the pixel block k;
computing, for each pixel block k, a standard deviation for luma pixel values included in the pixel block k;
determining, for each pixel block k, a block-mean bin index index b k , wherein the entire HDR luma codeword range is divided into N B non-overlapping luma bins, wherein each luma bin b represents a corresponding discrete luma codeword range, and wherein the block-mean bin index b k is the floor function of multiplying the mean for luma pixel values included in the pixel block k with N B ;
computing, for each luma bin b, a block histogram by counting a number of pixels that have a block-mean bin index b k equal to b;
computing, for each luma bin b, a block standard deviation by averaging standard deviation over all pixel blocks that have a block-mean bin index b k equal to b;
determining, for each luma bin b, a banding risk value, wherein the banding risk value is determined using a multiplicative combination of the block histogram and the block standard deviation of the respective luma bin;
determining a Gaussian function defined by Gaussian function parameters including a mean value and a width value of a corresponding Gaussian distribution, wherein the width value is the reciprocal of the doubled square of the standard deviation of the Gaussian distribution, wherein the mean value of the Gaussian function parameters is set to the maximum banding risk value across all luma bins;
generating a differential reshaping function using the Gaussian function parameters, the differential reshaping function specifying the amount of increment to the next lower input value of the differential reshaping function;
computing a luma-based forward reshaping function based on the differential reshaping function, wherein the luma-based forward reshaping function is a monotonically non-decreasing function that transfers a higher bit-depth codeword to a lower bit-depth by using the differential reshaping function; and
generating an output image for each image frame by applying the luma-based forward reshaping function to the respective image frame.
8. The method according to claim 7 , wherein the width value of the Gaussian function parameters is selected based on image statistics such that that the most banding prone luminance range is covered by the Gaussian function.
9. The method according to claim 7 , further comprising:
storing a normalized base-layer codeword as a single-channel forward reshaping function.
10. The method of according to claim 7 , further comprising:
determining a backwards reshaping function based on the luma-based forward reshaping function.
11. The method according to claim 10 , wherein the backwards reshaping function is approximated in the form of an 8-piece 1 st order polynomial curve.
12. The method according to claim 7 , wherein the differential reshaping function defines a number of codewords allocated to a given luminance range.
13. A non-transitory computer-readable medium storing instructions that, when executed by an electronic processor, cause the electronic processor to perform operations comprising the method according to claim 1 .