CPR Video Coding Buffer Update via Separate HMVP Tables
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current Picture Referencing (CPR) coding techniques in video encoding are inefficient for camera-captured video contents due to difficulties in finding exact matches within the same picture, especially when textual patterns have gradual changes over space, limiting coding performance.
Innovation Solution
Implementing two separate History-based Motion Vector Predictor (HMVP) tables, one for normal inter prediction and another for CPR mode, with specific handling of motion information and reference frame indices to enhance coding efficiency by utilizing the current picture as a reference for prediction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If Current Picture Referencing (CPR) coding is used for camera-captured video contents, then coding efficiency should be improved, but finding exact matches within the same picture becomes difficult due to gradual changes in textual patterns over space
Solution Approach 1:
The patent divides the current picture into multiple blocks and processes each block independently with its own motion vector prediction. This segmentation allows the encoder to find local matches within smaller regions rather than requiring exact matches across the entire picture, thereby resolving the contradiction between coding efficiency and match accuracy for camera-captured content with gradual changes.
Solution Approach 2:
The patent introduces a block vector (BV) as a new parameter to replace traditional motion vectors for CPR mode. This parameter change enables more flexible reference to locations within the current picture, allowing the system to adapt to gradual changes in textual patterns while maintaining coding efficiency through the BV-based prediction mechanism.
2Device complexity
If a single HMVP table is used for both normal inter prediction and CPR mode, then device complexity is reduced, but coding performance deteriorates due to inability to handle different prediction modes separately
Solution Approach 1:
The patent segments the single HMVP table into two separate tables: one for normal inter prediction and another for CPR mode. This segmentation allows each table to be optimized for its specific prediction mode, with the CPR-specific table storing block vectors and reference indices tailored for current picture referencing, thereby improving coding performance without significantly increasing overall device complexity.
Solution Approach 2:
The patent implements dynamic selection between the two HMVP tables based on the prediction mode being used. The encoder and decoder can switch between tables as needed, allowing the system to adapt to different coding scenarios dynamically. This dynamic approach enables optimal performance for both normal inter prediction and CPR mode while keeping the implementation flexible and manageable.
Data Source
AI summary
Video processing methods comprise receiving input data of a current block in a current picture having a plurality of Virtual Pipeline Data Units (VPDUs), wherein each VDPU is a fixed size, performing a decoding process to decode Coding Units (CUs) in the current VPDU and generate decoded samples of the current VPDU, wherein each CU coded in a Current Picture Referencing (CPR) mode is decoded by fetching decoded samples in a reference block stored in a CPR referring buffer, and updating the CPR referring buffer by storing the decoded samples of the current VPDU when finishing the decoding process of the current VPDU, wherein the CPR referring buffer is updated one VPDU by one VPDU.