Research On Image Enhancement Algorithm Based On Retinex Theory

In the field of image enhancement, there have been many practical algorithms, such as histogram equalization algorithm, frequency smoothing filtering algorithm, etc. Retinex algorithm is one of important image enhancement algorithms. By analyzing the visual processing of the retina and the cortex, Retinex algorithm estimates and eliminates the background light signal of image through mathematical processing, resulting in a more clear image. This algorithm can effectively remove the light signal of image, and the image after enhancement will be closer to its real color, which is more conducive to extraction and analysis of key information of image. The algorithm has a good enhancement effect for fog image and low brightness image, and has a very wide range of applications in the field of monitoring, medical image processing and remote sensing image processing. Therefore, researches for Retinex algorithm and promoting the development of the field of image enhancement have a high theoretical and practical significance.This paper researches on applications of Retinex theory in the image enhancement field. As to existing Retinex algorithms cannot effectively control SNR and cannot control color cast in low-light situations, an intensity self-adaptive Retinex algorithm is put forward, which proves effective in the simulation. Research work includes the following: analyzing and implementing various types of Retinex algorithms, discussing the advantages and disadvantages of common Retinex algorithms by simulation, and demonstrating that these algorithms have small scope of application, poor effect of enhancement and high operational costs; improving Retinex image enhancement algorithm and putting forward an intensity self-adaptive Retinex algorithm. Firstly, in this new algorithm the luminance component of the image will be extracted in the image preprocessing. After that, the color of low-light image will be corrected with Gray-World assumption and Perfect-Reflector theory, and then images will be processed with multi-scale Retinex algorithm in HSI color space. Simulation results show that the new algorithm is effective, as well as the simulation of specific application scenarios is carried out. The experimental analysis shows that the improved algorithm has a good effect on the color correction and a high SNR on low-light image enhancement.