Multiscale Retinex with Data-dependent Offset

As one of methods to improve the image quality, there is a method called multiscale retinex (MSR) which has been proposed by D.J. Jobson et al. In MSR, the reection components of an image are extracted and emphasized, and then the image with improved quality is obtained. This method is very useful and powerful especially for the visibility improvement of dark regions of the image. However, the resulting image tends to give us the unnatural impression because luminance components are removed, and the global contrast of the image is decreased in the processing. In this paper, a new MSR with a variable offset, which changes dependently on the local luminance information of the image, is proposed in order to overcome the disadvantage of the conventional MSR, and to further improve the image quality. Through the experiments, the effectiveness of the proposed method is illustrated.