Motion Vector Estimation Using Union of Search Areas
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video transcoding methods are inefficient due to high processing power and time delays caused by complex rate-distortion optimization in modern video encoding standards, especially when downsampling videos for smaller screens or devices, as they require full decoding and re-encoding processes.
Innovation Solution
The method involves downscaling full-resolution motion vectors and using residual transform coefficients to improve motion estimation accuracy and speed during transcoding, by modifying the rate-distortion cost function and employing a transcoder architecture that includes predictors for motion vector and partition mode initialization, which reduces computational complexity and enhances encoding performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If full-rate-distortion optimization is used during transcoding, then encoding quality is improved, but processing time and computational complexity increase significantly
Solution Approach 1:
The patent applies preliminary action by performing motion estimation at full resolution before downsampling. The transcoder extracts motion vectors from the full-resolution encoded video, then uses these pre-computed motion vectors to guide the motion compensation process at downsampled resolution. This preliminary motion estimation eliminates the need to perform time-consuming full-rate-distortion optimization during the actual transcoding process, thereby reducing processing time while maintaining encoding quality.
2Manufacturing precision
If full decoding and re-encoding is performed, then comprehensive optimization is achieved, but computational complexity and processing power requirements increase
Solution Approach 1:
The patent applies the extraction principle by separating and reusing the motion estimation results from the original encoding process. Instead of performing complete decoding and re-encoding, the transcoder extracts the motion vectors and other relevant data from the encoded bitstream, then directly utilizes these extracted elements for motion compensation at the target resolution. This selective extraction and reuse of encoding elements significantly reduces computational complexity while maintaining comprehensive optimization benefits.
Solution Approach 2:
The patent applies copying by creating a downsampled version of the reference frames and current frames, then performing motion compensation on these copied downsampled versions using the original motion vectors. This copying approach allows the transcoder to work with reduced-resolution data throughout the encoding process, thereby reducing computational complexity and processing requirements while still achieving optimized encoding results for the target resolution.
3Measurement precision
If motion vector search area is expanded to improve accuracy, then motion vector selection accuracy is improved, but processing time increases
Solution Approach 1:
The patent applies local quality by adapting the motion vector search strategy to the specific characteristics of each video region and motion type. Instead of uniformly expanding the search area across the entire frame, the transcoder uses the extracted motion vectors to identify regions with significant motion and concentrates the refined search in those local areas. This localized approach to motion vector refinement maintains high selection accuracy for important motion regions while avoiding unnecessary processing time spent on static or low-motion areas.
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 each downscaled motion vector pinpoints a search area within a reference frame. The union or combination of search areas defines the search field for candidate motion vectors. A motion vector is selected from the candidates based on realizing a minimum rate-distortion cost.