Motion Vector Merge Prediction for Faster Video Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies face challenges in optimizing encoding and decoding processes to efficiently handle increasing amounts of digital video data, particularly in selecting appropriate motion vectors and correction processing for prediction images.
Innovation Solution
An encoder and decoder derive an average value of motion vector values and register it as new motion vector information, along with correction processing information, to enhance the prediction candidate list, allowing for improved selection of suitable motion vectors and correction processing for each block, thereby increasing coding and decoding efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If conventional motion vector selection methods are used in H.265/HEVC, then the encoding process is simpler, but coding efficiency deteriorates due to inability to select optimal motion vectors from expanded candidate lists
Solution Approach 1:
The patent pre-calculates and stores correction processing information (such as gradient values, offset values, and correction flags) along with motion vector candidates in the candidate list. This preliminary preparation allows the encoder to quickly access and apply appropriate correction processing when selecting motion vectors, improving coding efficiency without significantly increasing encoding complexity during the actual encoding process.
Solution Approach 2:
The patent introduces correction processing information as an intermediary element that mediates between the motion vector candidate selection and the final prediction image generation. This intermediary contains pre-computed correction data (gradient values, offset values) that facilitates more accurate prediction without requiring complex real-time calculations, thus resolving the contradiction between coding efficiency and encoding complexity.
2Manufacturing precision
If more prediction candidates are added to the candidate list, then the likelihood of selecting optimal motion vectors increases, but the processing time and computational load increase
Solution Approach 1:
The patent performs preliminary calculations of correction processing information (gradient values, offset values, correction flags) for all motion vector candidates during the candidate list construction phase. This pre-computation allows the encoder to quickly evaluate and select from an expanded candidate list without incurring significant additional processing time during the actual encoding, as the heavy computational work is already done in advance.
Solution Approach 2:
The patent dynamically adjusts the candidate list by inserting new motion vector candidates (such as average motion vectors from neighboring blocks) and associating them with pre-computed correction processing information. This dynamic expansion of the candidate list with pre-prepared correction data increases selection accuracy while maintaining processing efficiency, as the system adapts the candidate list content without requiring exhaustive real-time computations for each candidate.
3Manufacturing precision
If correction processing is applied to all prediction candidates, then prediction accuracy improves, but encoding complexity and processing load increase
Solution Approach 1:
The patent applies correction processing selectively to specific regions or blocks based on local characteristics. The correction processing information includes correction flags and region-specific gradient values that allow the encoder to apply correction processing only where it is most beneficial (e.g., at block boundaries or in regions with significant illumination changes), rather than uniformly to all prediction candidates. This localized approach improves prediction accuracy in critical areas while minimizing the overall encoding complexity and processing load.
Data Source
AI summary
An encoder includes circuitry and memory connected to the circuitry. In operation, the circuitry: derives an average value of motion vector values of two prediction candidates in a prediction candidate list for a merge mode, and registers the average value derived as new motion vector information of a new prediction candidate into the prediction candidate list; and derives new correction processing information regarding correction processing of a prediction image, and registers the new correction processing information derived into the prediction candidate list in association with the new motion vector information.