Video Image Dehazing Algorithm Based on Multi-scale Retinex with Color Restoration

Abstract

Intelligent video surveillance in modern city is becoming more and more popular, and higher quality of images is required in video automatic analysis and recognition. In this paper, we develop a kind of haze-free method for outdoor foggy video image with low contrast. The method is based on the multi-scale guided filtering combined with Retinex algorithm. Compared with the traditional series method of Retinex, it may better in the following aspects: process the details in image enhancement, color fidelity and overcome noise etc. The experimental results show the advantages of our algorithm through three indicator values: information entropy, average gradient, and variance.