Point-Asymmetric Filter for Accurate Distance Measurement
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image processing techniques face challenges in achieving highly accurate distance measurement due to the use of different point spread functions for multiple-viewpoint images, which complicates the calculation of phase differences and results in inaccurate blur correction.
Innovation Solution
An image capturing apparatus with a filter that changes the point spread function of sensor images into point-asymmetric forms, using a sensor controller and distance calculator to convolute blur kernels and calculate the distance to an object based on correlations between the target and reference images, thereby correcting blur shapes and achieving accurate distance estimation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If different point spread functions are used for multiple-viewpoint images in conventional techniques, then the blur correction can be performed, but the distance measurement precision deteriorates
Solution Approach 1:
The patent applies asymmetry by using a point-asymmetric filter region in the aperture that creates different point spread functions for different color components (R, G, B) while maintaining a consistent asymmetric pattern across all color channels. This asymmetric filter design enables the calculation of phase differences between multiple-viewpoint images with the same sign, thereby achieving both accurate blur correction and precise distance measurement simultaneously
2Ease of operation
If conventional blur correction techniques are applied to multiple-viewpoint images, then the images can be processed, but the calculation of phase differences becomes complex and inaccurate
Solution Approach 1:
The patent implements local quality by applying different transmission factors to different regions of the aperture for different color components. The point-asymmetric filter region is positioned such that it affects R, G, and B components differently, creating local variations in point spread functions that simplify phase difference calculation while maintaining processing ease
Data Source
AI summary
An image capturing apparatus includes a sensor controller and a distance calculator. The sensor controller acquires a target image and a first reference image. The point spread function (PSF) of the target image is point-asymmetric, and the PSF of the first reference image being point-symmetric. The image sensor receives light having passed through a filter region that changes PSFs for sensor images of at least one kind into point-asymmetric forms, and then generates the target image for which a PSF has been changed into a point-asymmetric form by the filter region, and the first reference image that is not the target image. The distance calculator calculates the distance to an object captured in the target image and the first reference image, in accordance with the correlations each between an image obtained by convoluting a blur kernel to the target images and the first reference image.