针对MSR(multi-scale Retinex)算法对雾天彩色图像增强时，不能同时有效地保持细节增强与色彩保真的问题，通过将基于小波变换域的信息融合策略取代MSR算法中线性加权策略，提出一种新的MSR改进算法。融合的基本思路是：首先将待融合图像经四层小波分解，然后将高频分量取绝对值最大值来突出图像的细节，低频分量采用基于局部方差的技术调节色彩，实现保真的效果。通过主观观测和客观评价表明：在对雾天图像进行增强时，MSR改进算法比传统的MSR算法在细节增强和颜色保真方面具有更好的效果。

A novel multi-scale Retinex method is proposed to improve traditional Multi-scale Retinex in this paper, mainly aiming to solve the problem of not being able to maintain the details and the color fidelity effectively, by adopting information fusion strategy based on wavelet transform domain to replace the linear weighted average. At first, the basic strategy of image fusion is taking four wavelet decomposition, then the high-frequency take absolute maximum to enhance the details and the low-frequency adopt the strategy which based on local variance to modify color. The subjective and objective evaluation show that this MSR algorithm is better than the traditional MSR method in details and color fidelity in fog image enhancement.