Image Recovery Method Using Local Window Radiance Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional dehazing methods fail to accurately recover images due to rough transmission estimates, especially in small-scale areas, and assume identical attenuation coefficients for color channels, leading to inaccurate radiance representation.
Innovation Solution
An image recovery method that defines local windows around each pixel, assuming identical original radiance within the window to derive pixel-specific transmission maps with a global optimal solution, and allows different attenuation coefficients for color channels to correct for environmental medium effects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If conventional dehazing methods use assumptions of latent variables on single image, then the processing can be implemented on single image, but the transmission estimate becomes rough and inaccurate especially in small-scale areas
Solution Approach 1:
The patent divides the image into multiple local windows (e.g., 3x3 pixels centered at each pixel) and performs minimization estimation independently for each local window. This segmentation approach allows the method to capture fine-grained transmission variations in small-scale areas while maintaining single-image processing capability, resolving the contradiction between ease of operation and measurement precision.
2Device complexity
If conventional methods assume identical attenuation coefficients for all color channels, then the processing is simplified, but the radiance recovery becomes inaccurate
Solution Approach 1:
The patent assigns different attenuation coefficients to different color channels (red, green, blue) based on their respective wavelengths and haze interaction characteristics. This local quality differentiation allows accurate radiance recovery for each color channel while maintaining reasonable processing complexity through the use of standard minimization estimation algorithms.
Data Source
AI summary
An image recovery method is disclosed for eliminating an effect of an environmental medium. The image recovery method includes receiving a captured image affected by the environmental medium, defining a respective local window with each pixel located at a center of the respective local window, and under an assumption that original radiance of all pixels within the respective local window in the captured image are identical, performing a minimization calculation for each local window, to obtain a corresponding transmission of each pixel in the captured image.