Method and apparatus for adaptive context modeling in video encoding and decoding

A method for video encoding, comprising:

• accessing a quantized transform coefficient of a block of a picture in a video;
• obtaining a probability of said quantized transform coefficient being zero or non-zero, based on statistics of transform coefficients of at least a previous picture, wherein said probability is obtained further responsive to a scaling factor that is based on a ratio of a quantization parameter of said at least a previous picture and a quantization parameter of said picture;
• determining a context model for a syntax element associated with said quantized transform coefficient, responsive to said probability and a position of said quantized transform coefficient;
• and
• entropy encoding said syntax element based on said context model.

An apparatus for decoding video data, comprising at least a memory and one or more processors, wherein said one or more processors are configured to:

• obtain a probability of a quantized transform coefficient, of a block of a picture, being zero or non-zero, based on statistics of transform coefficients of at least a previous picture, wherein said probability is obtained further responsive to a scaling factor that is based on a ratio of a quantization parameter of said at least a previous picture and a quantization parameter of said picture;
• determine a context model for a syntax element associated with said quantized transform coefficient, responsive to said probability and a position of said quantized transform coefficient;
• entropy decode said syntax element based on said context model;
• and
• decode said quantized transform coefficient responsive to said decoded syntax element.

An apparatus for encoding video data, comprising at least a memory and one or more processors, wherein said one or more processors are configured to:

• access a quantized transform coefficient of a block of a picture in a video;
• obtain a probability of said quantized transform coefficient being zero or non-zero, based on statistics of transform coefficients of at least a previous picture, wherein said probability is obtained further responsive to a scaling factor that is based on a ratio of a quantization parameter of said at least a previous picture and a quantization parameter of said picture;
• determine a context model for a syntax element associated with said quantized transform coefficient, responsive to said probability and a position of said transform coefficient;
• and
• entropy encode said syntax element based on said context model.