Stereo-Image Disparity Quality Assessment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current quality measurement techniques for stereoscopic content, such as Peak Signal Noise Ratio (PSNR), are ineffective for 3D content and often rely on time-consuming subjective methods, which are not straightforward to analyze, especially for live content where accurate disparity computation is challenging.
Innovation Solution
A method that generates a sparse disparity map to assess stereo-image quality based on disparity range and rate of change, allowing for adjustments to disparity values to improve quality, considering factors like display type, size, and user preference, using techniques like SIFT feature matching and median filtering.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If subjective quality measures are used to assess stereo-image quality, then measurement effectiveness is improved, but time consumption increases and analysis becomes more complex
Solution Approach 1:
The patent replaces subjective human evaluation (mechanical process) with automated objective metrics that compute quality based on disparity map characteristics. The system automatically calculates metrics like disparity range, mode disparity, and histogram statistics to assess quality without requiring human observers, thereby reducing time consumption while maintaining measurement effectiveness.
Solution Approach 2:
The system performs self-assessment by automatically analyzing its own output (disparity maps) to determine quality metrics. The automated quality measurement system evaluates its own performance using computable characteristics of the disparity data, eliminating the need for external subjective evaluation and reducing both time and complexity.
2Measurement precision
If dense disparity computation is performed for all features, then measurement precision is improved, but processing time and computational complexity increase
Solution Approach 1:
The patent segments the disparity computation process by selecting and analyzing only a subset of salient features rather than processing all pixels uniformly. The system identifies and focuses on distinctive features (corners, edges, texture regions) that provide the most informative disparity data, reducing computational load while maintaining precision for quality assessment.
Solution Approach 2:
The system applies different processing strategies to different regions of the image based on their importance. Instead of uniform processing, the patent weights and prioritizes salient features locally, applying more computational resources to regions with distinctive characteristics while reducing processing in homogeneous areas, thereby optimizing the balance between precision and efficiency.
Data Source
AI summary
A variety of implementations are described. At least one implementation modifies one or more images from a stereo-image pair in order to produce a new image pair that has a different disparity map. The new disparity map satisfies a quality condition that the disparity of the original image pair did not. In one particular implementation, a first image and a second image that form a stereo image pair are accessed. A disparity map is generated for a set of features from the first image that are matched to features in the second image. The set of features is less than all features in the first image. A quality measure is determined based on disparity values in the disparity map. The first image is modified, in response to the determined quality measure, such that disparity for the set of features in the first image is also modified.