关键词: 低照度图像, 图像恢复, 图像增强, 带色彩恢复的多尺度Retinex算法（MSRCR）, MSRCP算法

Abstract: In the field of video surveillance and scene restoration, low illumination image has many noise points, low brightness and poor visual effect. However, the existing image processing technology is prone to color distortion and serious halo color block. In order to solve this problem, according to Weber-Fisher’s law, the pixels of the image are transformed into logarithmic space, and the preliminary enhanced image corresponding to the characteristics of the visual system is obtained adaptively. The average Gauss filtering results of the R channel of the image are calculated respectively to obtain the estimation of incident light. The difference between the pixel value of the preliminary enhanced image and the estimated incident light is taken as the result image of the multi-scale retinal algorithm. The RGB channel of the result image is mapped to the range of 0-255 according to the color proportion of the primary enhanced image. The final output image is obtained by fusing the three channels. Compared with MSR, MSRCR and MSRCP, the proposed algorithm can acquire the enhanced image adaptively according to the scene of different low illumination images, and is superior to MSR, MSRCR and MSRCP in contrast and chroma preservation. Experiments show that the algorithm has a good effect in low illumination image enhancement, restoration and chromaticity preservation, and has a good value for application in enhancing the effectiveness of video surveillance.

Key words: low illumination image, image restoration, image enhancement, Multi-Scale Retinex with Color Restoration（MSRCR）, MSR with Chromaticity Preservation（MSRCP）