Improvement of Contrast and Haloing Artifact for Single-Scale Retinex Model

Abstract

A digital camera is hard to catch luminance change of a scene including highlights and shadows, while human vision has the ability to sense a very wide range of scene luminance. The Retinex algorithm is known as a human perception-based image processing algorithm which is useful for contrast improvement for the very wide range scene. The MSR(Multi-Scale Retinex)algorithm previously proposed for suppressing halo artifacts, which integrates multiple SSR(Single-Scale Retinex), has the problem of high computation cost. This paper proposes a fast Retinex algorithm without halo artifacts by compressing dynamic range of a luminance-surround field in the SSR and using the bilateral filtering.