Disparity Map Correction Using Weighted Average
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing disparity map generation methods using histograms result in approximate disparity values due to bin width requirements, leading to incorrect judgments of continuous 3D objects as discontinuous.
Innovation Solution
A disparity map generation apparatus that uses a weighted average method based on the physical properties of objects from different viewpoints, determining weights for surrounding pixels to correct pixel disparities, preventing rough disparity values.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a histogram-based correction method is used, then disparity correction can be performed, but the disparity values become approximate and rough due to bin width requirements
Solution Approach 1:
The patent changes the fundamental parameter of the correction approach from histogram-based discrete binning to a continuous weighted average method. Instead of forcing disparity values into fixed bins, the invention uses weights based on pixel intensity similarity and spatial distance to continuously adjust disparities, thereby maintaining precision without the roughness introduced by bin width constraints
Solution Approach 2:
The patent creates a weight map that copies and adapts the disparity information from surrounding pixels to the target pixel. By copying disparity values from analogous pixels (those with similar intensity profiles) and weighting them by spatial distance and intensity similarity, the method preserves continuous disparity information rather than discretizing it into bins
2Ease of operation
If bin width is increased to determine mode values, then histogram-based correction can be performed, but the disparity values become more approximate and less precise
Solution Approach 1:
The patent inverts the conventional approach by not forcing continuous disparity values into discrete bins. Instead of binning disparities and finding modes, the invention starts with continuous disparity values and applies continuous weighted averaging, thereby inverting the discretization process and preserving precision throughout the correction operation
3Reliability
If histogram mode values are used for correction, then disparity correction can be achieved, but continuous 3D objects may be judged as discontinuous
Solution Approach 1:
The patent incorporates feedback mechanisms by using the disparity values from all pixels in the correction area to compute a weighted average for each pixel. The weights are determined by feedback from intensity similarity measurements and spatial distance calculations, allowing the system to adaptively correct disparities while maintaining the continuity of 3D objects through consistent feedback-driven weighting
Data Source
AI summary
There is provided an apparatus for generating a disparity map. According to the apparatus, an image acquisition section acquires a right-camera image and a left-camera image which are picked up from mutually different viewpoints by cameras. A disparity map generation section generates a disparity map that includes disparity expressed in terms of a difference in pixel positions between each of pixels and a corresponding one of pixels in the right-camera image and the left-camera image, respectively, as acquired. A weight determination section determines a weight that serves as a degree of contribution to disparity correction, for each of weight-calculation-target pixels composed of surrounding pixels. The weight-calculation-target pixels include a correction-target pixel or surrounding pixels that are present around the correction-target pixel in the disparity map. A disparity map correction section corrects the disparity of the correction-target pixel, using weighted average based on the determined weight.