Inter-Prediction Index Coding for Robust Video Decoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The dependency of inter-prediction candidate index decoding on the construction of the corresponding candidate list in video coding leads to robustness and throughput issues, causing cycle time overhead and making real-time decoding difficult, and is prone to errors due to corruption in encoded motion data.
Innovation Solution
Decouple the encoding of inter-prediction candidate indices from the construction of the corresponding candidate lists by using a predetermined maximum size for the candidate lists, allowing independent decoding and context selection based on CABAC bin indices, and optionally adding virtual candidates to achieve the maximum size.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If inter-prediction candidate index decoding is dependent on the construction of the corresponding candidate list, then decoding accuracy can be maintained, but decoding throughput decreases and cycle time overhead increases
Solution Approach 1:
The patent segments the decoding process into two independent parts: (1) decoding the inter-prediction candidate index using truncated unary decoding with a predetermined maximum codeword size, and (2) constructing the candidate list. This segmentation allows the index decoding to proceed independently without waiting for candidate list construction, thereby improving decoding throughput while maintaining accuracy through the predetermined decoding parameters.
Solution Approach 2:
The patent applies preliminary action by establishing a predetermined maximum codeword size for truncated unary decoding before the actual decoding process. This pre-defined parameter allows the decoder to immediately decode the candidate index without needing to know the actual candidate list size in advance, reducing cycle time overhead and enabling real-time decoding while preserving decoding accuracy through the pre-established decoding framework.
2Measurement precision
If inter-prediction candidate index decoding depends on candidate list construction, then context selection can be accurate, but robustness against encoded motion data corruption decreases
Solution Approach 1:
The patent extracts the decoding process from the candidate list construction process, allowing the inter-prediction candidate index to be decoded independently using a predetermined maximum codeword size. This extraction creates a robust decoding mechanism that does not depend on the actual candidate list contents, thereby maintaining reliability against motion data corruption while preserving context selection accuracy through the independent decoding path.
3Adaptability or versatility
If a variable size candidate list is used, then coding flexibility is improved, but device complexity increases due to dynamic list management
Solution Approach 1:
The patent changes the parameter of candidate list size from a dynamic variable to a predetermined maximum value for decoding purposes. This parameter change simplifies the decoding process by using a fixed maximum codeword size for truncated unary decoding, reducing device complexity while maintaining coding flexibility through the use of virtual candidates to achieve the maximum size when needed.
Data Source
AI summary
Methods are provided for inter-prediction candidate index coding independent of the construction of the corresponding inter-prediction candidate list, i.e., a merging candidate list or an advanced motion vector predictor list. A maximum allowed number of inter-prediction candidates for an inter-prediction candidate list is used for encoding the inter-prediction candidate index in an encoded bit stream. The maximum allowed number may be pre-determined or may be selected by the encoder and encoded in the bit stream. A decoder may then decode the index using the maximum allowed number of inter-prediction candidates independent of the construction of the corresponding inter-prediction candidate list.