Motion Vector Estimation Using SAD Thresholds for Video Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing motion vector estimation methods, such as those based on Sum of Absolute Differences (SAD), struggle to find unique motion vectors in images with repeated patterns, flat backgrounds, or overlapping objects, leading to deterioration in image quality, especially during frame rate up-conversion.
Innovation Solution
A method and apparatus that determine a motion vector by setting a local search area in another frame, comparing SAD values between reference pixel blocks and the current block, and selecting either the determined motion vector or a previous motion vector based on SAD difference values and threshold criteria, with adjustable allowable ranges and threshold values set according to the standard deviation of pixel values.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If SAD-based motion vector estimation method is used, then computational simplicity is maintained, but unique motion vector cannot be found in repeated pattern portions, flat background portions, and overlapping object portions
Solution Approach 1:
The patent divides the search space into multiple candidate regions and segments the evaluation process into multiple stages. Instead of evaluating all possible motion vectors in a large search area, the method segments the search into candidate motion vector identification, verification candidate selection, and final verification stages, making the complex evaluation manageable while ensuring unique motion vector detection
Solution Approach 2:
The patent performs preliminary actions by pre-identifying candidate motion vectors based on preliminary evaluation criteria before conducting the final verification. This preliminary action filters out obviously incorrect candidates early in the process, reducing computational complexity while ensuring that the unique motion vector is not missed in the final verification stage
2Measurement precision
If local search area is set to cover all possible motion vectors, then unique motion vector can be found, but computational complexity increases significantly
Solution Approach 1:
The patent applies local quality by setting different search ranges and evaluation criteria for different candidate motion vectors based on their likelihood of being correct. Instead of uniformly searching all possible motion vectors with the same effort, the method concentrates computational resources on promising candidates identified through preliminary evaluation, achieving accurate detection with reduced complexity
Solution Approach 2:
The patent performs partial action by evaluating only a selected subset of candidate motion vectors in detail rather than exhaustively evaluating all possible motion vectors. The method identifies and focuses computational effort on verification candidates that are most likely to be correct, achieving sufficient accuracy without the excessive computational cost of complete enumeration
3Measurement precision
If multiple candidate motion vectors are evaluated in detail, then unique motion vector can be identified, but processing time increases
Solution Approach 1:
The patent performs preliminary action by pre-screening motion vector candidates using fast evaluation criteria before conducting detailed verification. This preliminary filtering identifies verification candidates that are likely to be correct, allowing detailed evaluation to be performed on only a small subset of candidates, thus maintaining high identification accuracy while significantly reducing processing time
Solution Approach 2:
The patent applies skipping by rapidly eliminating unlikely motion vector candidates through quick preliminary evaluation without performing detailed verification. The method rushes through the evaluation of non-promising candidates using simplified criteria, focusing detailed analysis only on verification candidates, thereby reducing overall processing time while maintaining accuracy
Data Source
AI summary
A motion vector estimation method of a motion vector estimation apparatus are provided. The motion vector estimation method of the motion vector estimation apparatus includes: determining a motion vector of a current block by setting, in another frame, a local search area for searching a plurality of reference pixel blocks corresponding to the current block which is one of pixel blocks in a single frame; and finally determining one of the determined motion vector and a previous motion vector referred to in order to determine the local search area as a motion vector of the current block, based on sum of absolution difference (SAD) values of the reference pixel blocks included in the local search area.