Context Reduced Last Transform Coding for Video Entropy Encoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding techniques require a large number of contexts during parallel entropy encoding, leading to increased processing overhead, despite advancements in codecs like H.264/MPEG-4 AVC and JCT-VC standards.
Innovation Solution
The implementation of a context reduced last transform (CRLT) position coding technique, which shares context models among bins based on bin width, reducing the number of contexts needed for encoding the last transform position from 120 to 82 in YUV 4:2:0 video, by assigning all but the first three bins to share a context model with at least one other bin.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a large number of contexts are used for parallel entropy encoding of last transform position, then encoding precision is improved, but processing overhead increases
Solution Approach 1:
The patent merges multiple context models by having bins share context models. Specifically, bins are assigned to share context models with other bins, reducing the total number of context models from 120 to 82 for YUV 4:2:0 video, thereby reducing processing overhead while maintaining encoding precision
Solution Approach 2:
The patent makes context models universal by allowing them to serve multiple bins. A single context model is shared among multiple bins based on bin width criteria, enabling one context model to perform the function of multiple separate context models, thus reducing overall complexity
2Device complexity
If context models are shared among bins based on bin width, then processing overhead is reduced, but encoding precision may deteriorate
Solution Approach 1:
The patent applies local quality by differentiating context sharing based on bin width. Bins with width less than 4 have different context sharing behavior compared to bins with width 4 or greater, optimizing the balance between processing overhead and encoding precision for different bin characteristics
Data Source
AI summary
A context reduced last transform (CRLT) coding technique which enhances parallel context processing, such as utilized in JCTVC-D262, to reduce complexity by reducing the number of context models using for coding the position of the last significant transform coefficient. Selected context models are removed and additional bins are shared which reduce the number of contexts required. In one benchmark test for YUV 4:2:0 video, the number of context models were reduced from 120 for the proposed entropy encoding of JCTVC-D262 test model HM 2.0, versus 82 context models required for CRLT coding.