Segment-Based Disparity Decomposition for Stereo Imaging
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing stereo imaging algorithms face challenges in accuracy and computational efficiency, particularly in handling textured regions and semi-occlusions, and are prone to aliasing due to texture regions, despite advancements in GPU-based processing.
Innovation Solution
A novel surface/segment-based approach for disparity estimation using GPU-based heterogeneous sorting and residual compute stereo, which focuses on texture-based segmentation and disparity computation, and employs a Bayer pattern for demosaicing only the residual image, reducing computational load and improving accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional pixel-based or region-based stereo algorithms are used, then computational speed may be maintained, but accuracy deteriorates due to aliasing in textured regions and edge blurring
Solution Approach 1:
The patent divides the image into segments or regions with similar characteristics (color, texture, depth) and performs disparity computation on each segment rather than individual pixels. This segmentation approach reduces aliasing effects in textured regions while maintaining computational efficiency through region-based processing.
Solution Approach 2:
The patent applies different processing strategies to different regions of the image based on local characteristics. For example, textured regions are handled differently from smooth regions, and occluded regions receive special treatment. This local adaptation improves accuracy without uniformly increasing complexity across the entire image.
2Measurement precision
If segment-based approaches are used to improve accuracy, then disparity estimation precision improves, but computational load increases significantly
Solution Approach 1:
By segmenting the image into meaningful regions, the patent reduces the number of independent disparity computations required compared to pixel-based methods. Each segment is processed as a unit, significantly reducing computational load while maintaining or improving accuracy through coherent region-based matching.
Solution Approach 2:
The patent merges multiple pixels within a segment into a single representative value for disparity computation. This combining approach reduces the total number of computations required while preserving the essential visual information needed for accurate depth estimation.
3Measurement precision
If full image processing is performed to ensure complete analysis, then accuracy is maintained, but real-time performance requirements cannot be met
Solution Approach 1:
The patent extracts and processes only the most relevant features and regions for disparity computation, rather than analyzing every pixel in full detail. This selective extraction maintains accuracy for critical regions while reducing overall processing time to meet real-time requirements.
Solution Approach 2:
The patent applies partial processing strategies where full computational effort is devoted only to regions that require it (such as foreground objects or textured regions), while other regions receive simplified or skipped processing. This selective approach maintains accuracy where needed while achieving real-time performance overall.
Data Source
AI summary
A method and system for generating a disparity map. The method comprises the steps of generating a first disparity map based upon a first image and a second image acquired at a first time, acquiring at least a third image and a fourth image at a second time, and determining one or more portions comprising a difference between one of the first and second images and a corresponding one of the third and fourth images. A disparity map update is generated for the one or more determined portions, and a disparity map is generated based upon the third image and the fourth image by combining the disparity map update and the first disparity map.