Research On Low-light Color Image Enhancement And Denoising

Different from gray-scale images, color images contain more information, which can better meet people's visual perception, and help people as a basis for decision-making. With the boom of digital color image processing technology, it has successfully dominates in the area which were originally occupied by the gray-scale image processing technology and has the trend to replace the position. Certainly, the quality of the processed digital color image holds a special attraction. When the image is not suitable for viewing, some preprocessings are needed to improve the visual effects of the image. It is an important prerequisite of all subsequent processing, such as super-resolution reconstruction, target tracking, etc.Because of the dim lighting condition or being taken by backlight angle, the real photographs sometimes have poor image quality. Therefore, such images are often characterized by low brightness or contrast, large noise and small dynamic range. And solving these problems is our job. Retinex model is a proven effective way which can solve such problems. However, the application of traditional Retinex method might lead to the halo emergence and the effect of over enhancement. After description and analysis the Retinex model, a mean-shift filtering method is adopted to eliminate halo emergence, improved by our proposed method to reduce the effect of over enhancement. However, it needs to add limitations to the restraint function, or else leads to weak treatment. Then, this paper proposes a new histogram method, which re-enhanced the image after the part of Retinex model. The improved model can enhance the image quality and maintain the detail, while limits the restraint function. Experiment results show that the introduced methods do improve the system robustness. In addition, due to the serious color fading problem of dim-lightening color images, this paper applies the idea of Gamma correction into saturation channel to enhance the color, which gets the image better color information. Based on the analysis of the backlight compensation technology, a non parameter method is proposed in which the region of interest is not limited to the middle of the image any more. Actually, the proposed backlight compensation method further improves the robustness of the system. This article also analyzes the noise of low light image briefly, while denoises the image with the help of optical defocus model and block matching 3D filtering method.Finally, we make a summary and do a proper outlook for future work.