Current Picture Referencing with Adaptive Motion Vector Resolution
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding technologies face limitations in efficiently utilizing spatial correlation within a picture for screen content, particularly in finding exact matches for block-based prediction due to repetitive patterns, which hampers coding performance.
Innovation Solution
The implementation of current picture referencing (CPR) with adaptive motion vector resolution (AMVR) and sub-block prediction coding modes, enabling efficient syntax signalling to represent motion vectors and predict motion within the same picture, and disabling sub-block prediction when all reference pictures are the current picture.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If block-based prediction is used for screen content coding, then motion estimation can be performed, but exact matches for repetitive patterns cannot be found due to limitations in spatial correlation utilization
Solution Approach 1:
The current picture is divided into multiple current picture reference blocks that can serve as reference sources for motion estimation. This segmentation allows the encoder to search for exact matches of repetitive patterns within the same picture by utilizing multiple reference blocks, thereby improving match precision for screen content with repetitive structures.
Solution Approach 2:
The current picture serves dual functions: as the picture being encoded and as a reference picture for motion estimation. By enabling the current picture to function as its own reference (current picture referencing), the system improves adaptability for screen content coding while maintaining compatibility with traditional inter-picture prediction methods.
2Measurement precision
If adaptive motion vector resolution and sub-block prediction modes are implemented, then motion vector representation is improved, but syntax signalling complexity increases
Solution Approach 1:
The motion vector resolution is made adaptive rather than fixed. The system dynamically selects between different motion vector precisions (e.g., 1/4-pel, 1/8-pel, or integer pixels) based on the characteristics of the current block and reference blocks. This dynamic adaptation improves motion vector representation precision while avoiding the use of high precision where not needed, thus controlling syntax signalling complexity.
Solution Approach 2:
Different sub-blocks within a prediction unit can use different prediction modes and motion vector resolutions. This local quality approach allows high precision motion vectors to be applied only to regions requiring it, while other regions use simpler prediction methods, thereby improving overall motion representation without uniformly increasing syntax complexity across the entire block.
3Productivity
If current picture referencing is enabled with multiple reference blocks, then coding efficiency is improved, but the complexity of reference picture management increases
Solution Approach 1:
Reference blocks from the current picture are identified and prepared in advance before the actual motion estimation process. The encoder pre-processes the current picture to identify potential reference blocks and their positions, which simplifies the subsequent motion estimation step by having reference information ready, thus improving coding efficiency while managing complexity through advance preparation.
Data Source
AI summary
Methods and apparatus for a video coding system with the current picture referencing (CPR) mode enabled are disclosed. According to one method, when the current reference picture is equal to the current picture, an integer motion vector flag is inferred to be true without signalling the integer or parsing the motion vector flag. In another method, when all motion vector differences for the current block are equal to zero, an integer motion vector flag is inferred to be true without signalling the integer or parsing the motion vector flag. In yet another method, when all reference pictures for the current block are equal to the current picture: the sub-block prediction coding mode is disabled; and the current block is encoded or decoded by disabling the sub-block prediction coding mode. Alternatively, derived motion vectors associated with sub-blocks for the current block can be converted into integer motion vectors.