Subblock Coding Inference in Video Decoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies face inefficiencies in compressing video data, leading to increased storage needs and network bandwidth consumption, due to inaccurate inference rules for subblock flags in the Versatile Video Coding (VVC) standard.
Innovation Solution
The proposed solution involves improving the inference rules for subblock flags in video coding by determining the flag sb_coded_flag for a subblock based on conditions such as whether the subblock is a DC subblock or the last subblock with a non-zero coefficient level, and using these inferred flags to accurately estimate the context probability for entropy coding.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If existing inference rules for subblock flags are used, then video coding can be performed, but coding efficiency is reduced due to inaccurate context probability estimation
Solution Approach 1:
The patent modifies the inference rules for subblock flags by changing the parameters used to determine flag values. Specifically, it introduces new conditions based on transform coefficient levels and subblock positions (DC subblock, last subblock with non-zero coefficient) to accurately infer sb_coded_flag values. This parameter change in the inference logic directly improves context probability estimation accuracy, resolving the contradiction between maintaining coding functionality and improving coding efficiency.
2Device complexity
If transform skip is applied or transform skip residual coding is disabled, then coding complexity is reduced, but subblock flag inference becomes inaccurate
Solution Approach 1:
The patent applies preliminary action by establishing specific inference rules for subblock flags before the actual coding process. When transform skip is applied or transform skip residual coding is disabled, the patent pre-determines that certain subblock flags (sb_coded_flag) should be inferred as 0 based on the properties of DC subblocks and last subblocks with non-zero coefficients. This preliminary establishment of accurate inference rules ensures reliability is maintained while keeping coding complexity low.
3Measurement precision
If subblock flags are explicitly coded, then accuracy is maintained, but data size and transmission bandwidth increase
Solution Approach 1:
The patent extracts only the necessary subblock flag information that cannot be accurately inferred, and omits redundant flag coding. By using improved inference rules based on transform coefficient levels and subblock characteristics, the patent identifies which subblock flags can be reliably inferred (and thus don't need to be coded) versus which must be explicitly coded. This extraction approach maintains accuracy for essential flags while reducing overall data size by eliminating redundant flag transmissions.
Data Source
AI summary
A video decoder decodes a current transform block from a video bitstream. For a transform block encoded with regular residual coding, when a subblock flag sb_coded_flag is not present, the decoder infers the flag to be 1 if the subblock is a DC subblock and/or the last subblock in the transform block containing a non-zero coefficient level. Otherwise, sb_coded_flag is inferred to be 0. For a transform block encoded with regular residual coding, when the sb_coded_flag is present, the decoder determines a context index for an arithmetic decoding process based on the sb_coded_flag of previous subblocks and decodes the sb_coded_flag of the subblock using the arithmetic decoding process. The decoder decodes the transform block based on the determined sb_coded_flag.