Abstract: Aimed at the disadvantages of Retinex algorithm, an improved center surrounding function is proposed, and the conception of resolution for image enhancement is also presented. In order to take advantage of the complementary properties for color image enhancement between resolution and scale of Retinex, a Multi-Resolution and Multi-scale Retinex（MRMSR） algorithm for color image enhancement is proposed in this paper. In experiment, multiple color images are used to test and verify the novel algorithm, and the qualities of enhancement image are evaluated with subjective evaluation method and objective evaluation methods. Experimental results show, compared with MSR, MSRCR, etc., that the MRMSR not only has a better effect for image enhancement, and can effectively reduce the impactions of Retinex scales for image enhancement.

Key words: Retinex, improved center surrounding function, resolution of image enhancement, Multi-Resolution and Multi-scale Retinex（MRMSR）

摘要： 针对Retinex算法的缺点与不足，提出了一种改进的中心/环绕函数，以及图像增强分辨度概念。利用Retinex算法中图像增强分辨度与尺度之间对图像增强效果的互补特性，提出了多分辨多尺度的Retinex（Multi-Resolution and Multi-Scale Retinex，MRMSR）彩色图像增强算法。实验过程中，采用图像质量主观评价法和图像质量客观评价法相结合的方式，对多幅彩色图像进行了算法验证，并与MSR、MSRCR等算法进行比较。实验表明，MRMSR算法具有较好的图像增强效果，其图像增强效果明显优于其他算法，并能够有效地降低尺度对图像增强的影响。

关键词: Retinex, 改进的中心/环绕函数, 图像增强分辨度, 多分辨多尺度