Coding of residual data in predictive compression
View Patent ↗Encoding input data includes: generating a first block of coefficients based on a transform performed on a residual block of data for multiple pixels; generating reference information based on a reference block of data corresponding to the residual block of data; and determining losslessly decodable code values representing the first block of coefficients based on the reference information.
1. A method for encoding input data, the method comprising:
generating a first block of coefficients based on a transform performed on a residual block of data for multiple pixels;
generating reference information based on a reference block of data corresponding to the residual block of data; and
computing losslessly decodable code values representing the first block of coefficients based on the reference information, wherein computing losslessly decodable code values comprises determining portions of a code value representing respective portions of the first block of coefficients based on at least one value derived from at least a portion of the reference information, and wherein determining the portions of the code value comprises determining the portions based on respective estimated probabilities or codes determined according to the at least one value.
2. The method of claim 1 , wherein the estimated probabilities determined according to the at least one value are also determined according to one or more previously decoded code values.
3. The method of claim 2 , wherein the estimated probability for determining a first portion of the code value is based on a value stored in a data structure at a location identified by an index that depends on: a position within the first block of coefficients, and the at least one value.
4. The method of claim 3 , wherein the position within the first block of coefficients comprises a position in a one-dimensional ordering of the coefficients in the first block of coefficients.
5. The method of claim 3 , further comprising updating a value stored at a first location in the data structure based on: a value previously stored at the first location associated with the one or more previously decoded code values, and a value of a symbol representing a portion of the first block of coefficients.
6. The method of claim 2 , wherein the estimated probability for determining a first portion of the code value comprises a conditional probability that a symbol representing a portion of the first block of coefficients has a particular symbol value given the one or more previously decoded code values and the at least one value.
7. The method of claim 6 , wherein the symbol representing a portion of the first block of coefficients comprises a binary symbol determined according to a value of a coefficient in the first block of coefficients in a particular position with respect to a transform domain of the transform performed on the residual block.
8. The method of claim 1 , wherein the reference information based on the reference block of data comprises a second block of coefficients based on a transform performed on the reference block of data.
9. The method of claim 8 , wherein the at least one value according to which a particular probability is being estimated comprises a value based on a coefficient in the second block of coefficients that has a position within the second block of coefficients that corresponds to a position of a coefficient within the first block of coefficients for which the particular probability is being estimated.
10. The method of claim 8 , wherein the at least one value comprises a value based on a number of non-zero coefficients in the second block of coefficients.
11. The method of claim 8 , wherein generating the second block of coefficients based on a transform performed on the reference block of data comprises quantizing values resulting from the transform.
12. The method of claim 1 , wherein generating the first block of coefficients based on a transform performed on the residual block comprises quantizing values resulting from the transform.
13. The method of claim 1 , wherein the respective portions of the first block of coefficients comprise a series of symbols, with each symbol having a value determined by at least one coefficient of the first block of coefficients.
14. The method of claim 13 , wherein the symbols are binary symbols each having one of two possible values.
15. The method of claim 1 , wherein the respective portions of the first block of coefficients comprise a series of symbols, with each symbol having a value determined by at least one coefficient of the first block of coefficients.
16. The method of claim 15 , wherein the code value representing the respective portions of the first block of coefficients comprises an arithmetic code value generated based on the series of symbols and the respective estimated probabilities.
17. The method of claim 15 , wherein each of the series of symbols has a symbol value that is associated with a corresponding one of the respective estimated probabilities.
18. The method of claim 1 , wherein a set of codewords for losslessly decoding at least one of the code values is based on information stored in a data structure at a location identified by an index that depends on: a position within the first block of coefficients, and the at least one value.
19. The method of claim 18 , wherein the position within the first block of coefficients comprises a position in a one-dimensional ordering of the coefficients in the first block of coefficients.
20. The method of claim 1 , wherein the losslessly decodable code values are computed using a coding scheme selected from the group consisting of: arithmetic coding, and Huffman coding.
21. A non-transitory computer readable medium storing a computer program for encoding input data, the computer program including instructions for causing a computer system to:
generate a first block of coefficients based on a transform performed on a residual block of data for multiple pixels;
generate reference information based on a reference block of data corresponding to the residual block of data; and
compute losslessly decodable code values representing the first block of coefficients based on the reference information, wherein computing losslessly decodable code values comprises determining portions of a code value representing respective portions of the first block of coefficients based on at least one value derived from at least a portion of the reference information, and wherein determining the portions of the code value comprises determining the portions based on respective estimated probabilities or codes determined according to the at least one value.
22. An apparatus for encoding input data, the apparatus comprising:
a memory configured to buffer one or more frames reconstructed from the input data; and
at least one processor coupled to the memory and configured to process the input data based on the one or more frames buffered in the memory, the processing including:
generating a first block of coefficients based on a transform performed on a residual block of data for multiple pixels;
generating reference information based on a reference block of data corresponding to the residual block of data; and
computing losslessly decodable code values representing the first block of coefficients based on the reference information, wherein computing losslessly decodable code values comprises determining portions of a code value representing respective portions of the first block of coefficients based on at least one value derived from at least a portion of the reference information, and wherein determining the portions of the code value comprises determining the portions based on respective estimated probabilities or codes determined according to the at least one value.