Light Source Estimation Using Color Balance and Texture
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for estimating a light source in images, such as the brightness-color correlation method, are unstable when the image texture is flat and the number of colors is small, leading to inaccurate calculations and increased computational costs, especially in real environments with varying illumination conditions.
Innovation Solution
A light source estimating apparatus that calculates brightness-color correlation data from image frames, considers color balance of entire and neutral gray regions, and uses texture analysis and time sequence data to accurately determine the type of light source, reducing reliance on position and intensity of the light source and minimizing color feria.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If the brightness-color correlation method is used to estimate light source, then calculating cost is reduced, but estimation accuracy deteriorates in scenes with flat color textures and few colors
Solution Approach 1:
The patent combines multiple feature analysis methods (brightness-color correlation, color balance analysis, and texture analysis) into a unified light source estimation system. By merging these different approaches, the system achieves accurate light source estimation in diverse scenes while maintaining reasonable computational efficiency, resolving the contradiction between low calculating cost and high estimation accuracy.
Solution Approach 2:
The patent dynamically adjusts the weighting and selection of different feature parameters based on scene characteristics. When color texture is flat or colors are limited, the system shifts reliance from brightness-color correlation to color balance and texture features, thereby maintaining estimation accuracy across different scene types without requiring excessive computational resources.
2Measurement precision
If spectral characteristics of light source and object surface are used for estimation, then estimation accuracy is improved, but number of calculating operations increases
Solution Approach 1:
The patent extracts and utilizes only the most essential and easily obtainable features from image data (brightness, color balance, and basic texture statistics) rather than performing full spectral analysis. This extraction approach maintains sufficient estimation accuracy while dramatically reducing the number of calculating operations required, making the method suitable for real-time processing in digital cameras.
Solution Approach 2:
The patent implements a partial analysis approach where not all spectral characteristics are computed, but rather a selected subset of features that provide the most discriminative power for light source estimation. This partial action strategy achieves good estimation accuracy with significantly reduced computational complexity compared to complete spectral analysis.
3Ease of manufacture
If gray hypothesis is used for white balance adjustment, then calculating cost is low, but color feria occurs when average color deviates from gray due to scene content
Solution Approach 1:
The patent incorporates feedback mechanisms where the estimated light source type and scene texture characteristics are used to adjust and refine the white balance adjustment process. By continuously monitoring estimation results and scene features, the system can detect when gray hypothesis may lead to color feria and apply corrective adjustments, thereby improving white balance reliability while maintaining low calculating cost.
Solution Approach 2:
The patent dynamically changes the parameters and assumptions of the gray hypothesis based on detected scene characteristics. When texture analysis indicates flat color textures or when brightness-color correlation suggests potential color feria, the system adjusts the white balance calculation parameters accordingly, preventing color distortion while keeping computational costs low.
Data Source
AI summary
A method and apparatus for estimating a light source in an image obtained by an image device are provided. The light source estimating apparatus includes: a brightness-color correlation calculator which calculates brightness-color correlation data from an image of an object generated from at least one image frame; a color balance calculator which calculates a color balance of an entire region of the image and a color balance of a neutral gray region of the image having an average color of gray; a characteristic data generating unit which generates characteristic data of the image based on the brightness-color correlation data, the color balance of the entire region and the color balance of the neutral gray region; and an identifier which determines a type of a light source in the image, based on the characteristic data of the image.