The Research Of Color Image Enhancement Algorithm Based On Retinex

Due to the image is influence by rain, fog, light, and other a series of factors, the image will not be so clear,so the image not only can not meet the aesthetic needs of the human eyes, but will effect in later recognition and the feature extraction, So the image enhancement is an important technology of digital image processing. With the development of digital image technology, there are more and more images enhancement techniques have been proposed.In order to study better and more practical image enhancement algorithm, Land proposed the theory of Retinex to enhance image in 1963. Among them, Single-Scale Retinex (SSR), Multi-Scale Retinex (MSR) are good kinds in Retinex algorithm for image enhancement; And, in order to solve the problem of color distortion after image enhancement, Multi-Scale Retinex with Color Restoration (MSRCR) be put forward. Through analysis and comparison, find the following points can be improved:First, Retinex algorithm is easy to produce over enhancement, that is the enhanced image is obviously bright; Second, the image’s histogram is too narrow after be enhanced.Based on the above two points, puts forward the improved color image enhancement algorithm MSRCR. firstly it do denoising preprocessing, and then based on the histogram of the denoised images to adjust luminance in global to increase the proportion of dark pixels, and then use Retinex with color gray multi-scale enhance color image. Finally get the final enhanced image by linearly stretching based on the mean and variance of the enhanced image. Compare the image using enhancement algorithm with the image using other enhancement algorithm, it proves that this algorithm can achieve better effect of image enhancement by subjectively observing and objectively analyzing.