Confidence-Guided Disparity Refinement for Occlusion Artifacts
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing disparity estimation methods, particularly those using feature matching algorithms, produce imperfect disparity maps with errors and inaccuracies, such as artifacts near occlusions and missing regions, necessitating improved disparity refinement techniques.
Innovation Solution
A confidence-based fusion approach is employed, combining depth-from-stereo (DFS) disparity information with predicted disparity refinement information generated by a deep learning network, using a confidence map to weight the fusion of disparity maps at each pixel location, potentially incorporating an early termination mechanism to enhance accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If feature matching algorithms are used for disparity estimation, then depth information can be obtained from stereoscopic images, but the disparity maps contain errors and artifacts particularly near occlusions and missing regions
Solution Approach 1:
The patent introduces an intermediary deep learning network that processes the initial disparity map and confidence information to generate a refined disparity map. This intermediary system acts as a mediator between the raw feature matching results and the final output, filtering out errors and artifacts while preserving accurate disparity information.
Solution Approach 2:
The patent changes the parameter representation by introducing confidence information as an additional parameter that quantifies the reliability of each disparity estimate. This confidence parameter is then used to weight the refinement process, allowing the system to adaptively adjust the influence of different regions in the disparity map.
2Measurement precision
If confidence-based fusion is used to refine disparity maps, then accuracy is improved, but computational complexity increases due to the machine learning network processing
Solution Approach 1:
The patent performs preliminary action by generating confidence information from the initial disparity map before the refinement process. This pre-computed confidence map guides the subsequent refinement network, allowing it to focus computational resources on regions with low confidence rather than uniformly processing the entire image.
Solution Approach 2:
The patent applies local quality by using the confidence map to apply different refinement strategies to different regions of the disparity map. Regions with high confidence undergo minimal modification, while regions with low confidence receive more aggressive refinement, optimizing computational resources according to local needs.
Data Source
AI summary
Systems and techniques are provided for generating disparity information from two or more images. For example, a process can include obtaining first disparity information corresponding to a pair of images, the pair of images including a first image of a scene and a second image of the scene. The process can include obtaining confidence information associated with the first disparity information. The process can include processing, using a machine learning network, the first disparity information and the confidence information to generate second disparity information corresponding to the pair of images. The process can include combining, based on the confidence information, the first disparity information with the second disparity information to generate a refined disparity map corresponding to the pair of images.