Novel Color White Balance Algorithm

This MATLAB program implements a novel color white balance algorithm designed to address color cast phenomena in images captured under colored light sources. The algorithm presents an improved Laplacian of Gaussian (LoG) edge-based automatic white balance method that simultaneously applies LoG operators to both Cb and Cr components in YCbCr color space. The implementation leverages the zero-crossing characteristic of LoG edges to directly extract pixels on both sides of edges. To mitigate interference from dominant color textures, the algorithm incorporates edge suppression using color cast information from white regions. The extracted chromatic edges enable effective estimation of the original image's color cast, leading to calculation of gain factors for three color channels and subsequent white balance correction. The code structure includes key functions for color space conversion, LoG edge detection with zero-crossing identification, white region analysis for edge suppression, and channel gain computation. Extensive experiments conducted on numerous images demonstrate that this algorithm effectively performs white balance correction for images under various colored lighting conditions. Comparative experiments with multiple white balance algorithms confirm the method's broad adaptability and superior performance characteristics.