Hybrid Recursive Matcher for Real-Time Disparity Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current disparity estimation methods, such as the hybrid recursive matcher, are not capable of parallel processing and are limited by their meander-like recursive structure, making them inefficient for high-definition image resolutions and multi-core environments, which is essential for real-time disparity map generation in modern video processing applications.
Innovation Solution
The proposed solution involves a line-wise and column-wise hybrid recursive matcher (L-HRM) and temporal hybrid recursive matcher (T-HRM) that break the conventional meander recursion, allowing for parallel execution of lines and columns, respectively, enabling real-time disparity estimation suitable for implementation in hardware like FPGAs or ASICs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional hybrid recursive matcher with meander-like recursive structure is used, then disparity estimation can be performed, but parallel processing capability is lost and processing time increases
Solution Approach 1:
The patent segments the image processing into independent line-wise and column-wise operations. Each line and column can be processed independently and in parallel, breaking the sequential meander-like recursion into manageable segments that can be executed simultaneously on multi-core processors or hardware pipelines.
Solution Approach 2:
The patent introduces dynamic programming with memoization to replace the static meander-like recursive structure. This allows the algorithm to adaptively compute disparity values while maintaining parallel execution capabilities through independent line and column processing, rather than being constrained by a fixed recursive pattern.
2Loss of time
If meander-like recursive structure is used for disparity estimation, then algorithm can be implemented, but real-time processing capability is lost
Solution Approach 1:
By dividing the disparity estimation into independent line-wise and column-wise passes, the patent enables real-time processing through parallel execution. Each segment (line or column) can be processed independently without waiting for other segments, dramatically reducing total processing time compared to sequential meander-like recursion.
Solution Approach 2:
The patent performs preliminary disparity estimation along lines and columns before combining results. This preliminary action allows each segment to be pre-processed independently and in parallel, with final integration happening afterward, thus enabling real-time performance while maintaining algorithmic simplicity.
3Adaptability or versatility
If conventional disparity estimation method is used, then single-core processing is sufficient, but multi-core environments cannot be utilized effectively
Solution Approach 1:
The patent segments the processing task into independent line-wise and column-wise operations that can be distributed across multiple cores. Each core can independently process specific lines or columns simultaneously, fully utilizing multi-core environments to improve processing efficiency while maintaining the same disparity estimation quality.
Data Source
AI summary
An apparatus for estimating a disparity map based on at least two images is provided. The apparatus includes at least two processing units, which include a pixel recursion unit configured to determine a disparity value as a pixel recursion disparity candidate based on a plurality of pixel values of the at least two images and a selector configured to select a selected disparity candidate to determine at least one of the disparity map values of the disparity map. The selector is adapted to select the selected disparity candidate from a candidate group assigned to the selector. The candidate group assigned to the selector includes the pixel recursion disparity candidate, a second disparity candidate and a third disparity candidate. Moreover, the selector is adapted to select the selected disparity candidate independently from a different selector of a different processing unit of the at least two processing units.