Quantization matrix compression in video coding
AssigneeTexas Instruments Incorporated
InventorsMinhua Zhou, Vivienne Sze
Life of this patent
Estimating the term…
Abstract
A method of quantization matrix compression in a video encoder is provided that includes preprocessing a quantization matrix by performing at least one selected from down-sampling the quantization matrix and imposing 135 degree symmetry on the quantization matrix, performing zigzag scanning on the pre-processed quantization matrix to generate a one dimensional (1D) sequence, predicting the 1D sequence to generate a residual 1D sequence, and coding the residual 1D sequence using kth order exp-Golomb coding to generate a compressed quantization matrix, wherein k≥0.
Claims20
A method of decoding an encoded video bit stream, comprising:
- decoding the encoded video bit stream into an entropy decoded bit stream including decoded bits, the decoded bits comprising quantized transform coefficients;
- extracting a first flag from the decoded bits, the first flag indicating a non-default quantization matrix;
- extracting a second flag from the decoded bits, the second flag indicating a relationship between a first and second index position where the non-default quantization matrix will be implemented;
- and
- performing an inverse quantization operation on the quantized transform coefficients using the non-default quantization matrix at either the first or second index position.
A system for decoding an encoded video bit stream, comprising:
- an entropy decoding component configured to entropy the encoded video bit stream into decoded bits, the decoded bits comprising quantized transform coefficients;
- and
- an inverse quantization component configured to:
- extract a first flag from the decoded bits, the first flag indicating a non-default quantization matrix;
- extract a second flag from the decoded bits, the second flag indicating a relationship between a first and second index position where the non-default quantization matrix will be implemented;
- perform an inverse quantization operation on the quantized transform coefficients using the non-default quantization matrix at either the first or second index position.
A method comprising:
- decoding, by a decoder, an encoded video bit stream into quantized transform coefficients;
- determining, from a first indicator in the encoded video bit stream, the quantized transform coefficients are associated with a non-default quantization matrix;
- determining, from a second indicator in the encoded video bit stream, a reference quantization matrix based on a difference between a first scaling list index of the non-default quantization matrix and a second scaling list index of the reference quantization matrix;
- and
- performing an inverse quantization operation on the quantized transform coefficients based on the non-default quantization matrix and the reference quantization matrix.
Dependencies are inferred from the claim text.