Image enhancement using a nonlinear method with an improved Single-Scale Retinex algorithm

摘要

In this paper a new approach is proposed to enhance images that inspired by adaptive global and local processing of the human visual system. The method first uses a principal component analysis to the original image to provide orthogonality between channels and thus reduces the chromatic changes induced by the processing of luminance. Then a nonlinear method is applied to adjust the global dynamic range. Finally with an improved Sing-Scale Retinex algorithm whose filter shape is adapted to high-contrast edges of the image, the local contrast is increased without generating halo artifacts or washed-out problems. Our method is compared with other algorithms and renders more visually pleasing images.