Stereo Disparity Estimation via Segmented Scan Order
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional stereo vision techniques require significant computational resources and memory for disparity optimization, making real-time processing challenging due to large computation and memory burdens.
Innovation Solution
A disparity estimation method that optimizes scan order and uses energy functions based on edge detection to reduce computation time and memory usage, employing a strip-based scan order to improve resource efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional disparity optimization is performed by loading all cost values into memory and processing from left to right, then complete disparity optimization can be achieved, but computation time and memory usage become excessively large
Solution Approach 1:
The patent divides the image processing into multiple passes (first pass and second pass) with different scan orders. The first pass processes rows from top to bottom, and the second pass processes columns from left to right. This segmentation allows the algorithm to achieve complete disparity optimization without loading all cost values into memory simultaneously, thus reducing both computation time and memory usage while maintaining optimization completeness.
2Measurement precision
If all cost values are loaded into memory for disparity optimization, then accurate disparity calculation can be performed, but memory space requirements become excessively large
Solution Approach 1:
The patent segments the memory usage by processing the image in two separate passes with different scan orders. The first pass processes rows from top to bottom, and the second pass processes columns from left to right. This segmentation allows the algorithm to maintain accurate disparity calculation while using minimal memory space, as only the necessary cost values for the current processing step need to be retained in memory at any given time.
Solution Approach 2:
The patent performs preliminary processing in the first pass by calculating initial disparity values for each pixel row by row. These preliminary results are then used as input for the second pass, which refines the disparities by processing column by column. This preliminary action allows the algorithm to achieve accurate final disparities without needing to load all cost values into memory simultaneously.
3Reliability
If conventional scan order is used for disparity optimization, then complete image processing can be achieved, but the burden of real-time processing becomes heavy
Solution Approach 1:
The patent segments the processing into two distinct passes with different scan orders. The first pass processes rows from top to bottom to establish initial disparities, and the second pass processes columns from left to right to refine them. This segmentation enables real-time processing by reducing the computational burden on any single processing step while ensuring complete image processing is achieved.
Solution Approach 2:
The patent dynamically changes the scan order between passes. The first pass uses a row-wise scan order (top to bottom), while the second pass uses a column-wise scan order (left to right). This dynamic adjustment of processing direction allows the algorithm to efficiently utilize available computational resources and achieve real-time processing capability while maintaining processing completeness.
Data Source
AI summary
A disparity estimation method of stereoscopic image is provided. A matching cost computation is executed for a first image and a second image, and one extreme value is selected from cost values corresponding to estimating disparities for each pixel to obtain a matching point corresponding to each pixel. And a matching disparity corresponding to each matching point is adjusted based on edge detection according to a scan order.