Color Constancy-based Visibility Enhancement of Color Images inLow-light Conditions
基于颜色恒常性的低照度图像视见度增强

Color imaging in low-light conditions is often significantly degraded by insufficient lighting and color cast. In this paper, we present a novel color constancy-based method to enhance the visibility of low-light images. Due to the occurrence of the scene illuminant or specular highlights, inhomogeneous illumination affects the estimation of the illuminant color across the scene. Thus in the image, an active set of pixel values is defined to reduce their effects. We then propose an improvement on the gray-pixel assumption imposed by the shades-of-gray algorithm using the active set. The improved shades-of-gray algorithm is applied to low-light image visibility enhancement for cast removal. Limits of histogram clipping are found in the active set in the post-processing step that is added to render the resulting image with a good lightness and contrast range. Results on a variety of low-light images demonstrate that the proposed method can achieve a good correction for lightness, contrast and color fidelity without local graying-out artifacts or halo artifacts intrinsic to Retinex approaches.