Motion Vector Prediction in Video Transcoding Using Residual Weights
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video transcoding methods for spatial down-sampling are costly in terms of processing power and time due to the complexity of rate-distortion optimization in modern video encoding standards, and they often require full-resolution video decoding before down-sampling, which is resource-intensive.
Innovation Solution
The method involves downscaling full-resolution motion vectors, calculating weighting factors based on transform domain residual coefficients, determining motion vector predictions, and selecting desired motion vectors for encoding a downsampled video, leveraging full-resolution information to improve transcoding efficiency and accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If full-resolution video is decoded and then spatially downsampled using conventional transcoding methods, then the downsampled video can be generated, but the processing power consumption and time delay increase significantly due to rate-distortion optimization complexity
Solution Approach 1:
The patent applies preliminary action by performing motion vector prediction and residual calculation during the decoding phase of full-resolution video, before the actual spatial downsampling and encoding occur. This allows the transcoder to pre-compute motion information and exploit it during the downsampling process, avoiding the need to perform complex rate-distortion optimization on full-resolution data while maintaining encoding quality.
Solution Approach 2:
The patent extracts and utilizes only the essential motion vector information and residual data from the full-resolution video decoding process, rather than processing the entire full-resolution video stream through the complete encoding pipeline. This extraction approach reduces the computational burden by focusing only on the critical motion information needed for efficient downsampling.
2Manufacturing precision
If conventional rate-distortion optimization is employed in the transcoder, then encoding quality is improved, but the processing time and computational complexity increase
Solution Approach 1:
The patent uses copying by utilizing the motion vectors and residual data already available from the full-resolution video decoding process. Instead of performing complete rate-distortion optimization from scratch on the downsampled video, the system copies and adapts the motion information from the full-resolution source, reducing computational complexity while maintaining encoding quality through the exploitation of existing motion patterns.
Solution Approach 2:
Motion vector prediction and residual calculation are performed in advance during the decoding phase, before the downsampling and encoding processes. This preliminary computation of motion information allows the subsequent encoding to proceed with reduced complexity by using pre-computed motion models rather than performing exhaustive rate-distortion optimization.
3Measurement precision
If full-resolution video is decoded before down-sampling, then accurate motion information is available, but the processing resources and time required increase significantly
Solution Approach 1:
The patent performs motion vector prediction and residual calculation as preliminary actions during the decoding phase of full-resolution video. This allows accurate motion information to be obtained in advance, which then can be exploited during the downsampling process to achieve both high motion vector accuracy and reduced processing time by avoiding redundant computations.
Solution Approach 2:
The system extracts only the essential motion vector and residual information from the full-resolution video decoding process, rather than processing all pixel data through the complete encoding pipeline. This selective extraction maintains the accuracy needed for motion compensation while significantly reducing the time and resources required for processing.
Data Source
AI summary
A transcoder and methods of encoding inter-prediction frames of a downsampled video wherein the downsampled video is a spatially downsampled version of a full-resolution video. Full-resolution motion vectors are downscaled and a weighting factor is calculated for each downscaled motion vector based upon the transform domain residual coefficients associated with that full-resolution motion vector. A motion vector prediction is made based on the weighted average using the downscaled motion vectors and their weighting factors.