Research On Enhancement Algorithms Of Low Illumination Images And Videos Based On Retinex Theory

In everyday life,there will be short contrast,insufficient details and color deviation in images and videos because of the influence of environment or other factors.These not only affect the human's observation,but also affect the subsequent processing of such images and videos.Therefore,it has very important practical significance for enhancement of low-light images and videos.In this paper,the feature of low-light images and the existing algorithms are studied,and then the existing algorithms are improved and perfected according to these feature and the actual situation.The specific research contents of this paper are as follows:1.This paper presents a fusion-based image enhancement algorithm based on Retinex theory,which is mainly applied to low-light images.After studying many algorithms,this paper selects the guided filtering and the variational frame as the method of illumination component estimation,and we use the gradient fusion method to fuse the two illumination components and obtains the reflection component according to the Retinex theory.Then the illumination component and the reflection component are respectively nonlinearly stretched and gamma-corrected,after that the two components are combined into a new image.However,due to the possible over-enhancement,the result was finally processed by contrast-limited adaptive histogram equalization(CLAHE).The experimental results show that compared with the traditional algorithm,the proposed algorithm is more prominent in subjective aspect especially the whole brightness is obvious,and there is no obvious deviation in the color.On the objective aspect,the experimental results show the superiority of the proposed algorithm.2.This paper presents a fast enhancement algorithm based on video frame compensation.First,the single-frame image is estimated by gradient to extract the illumination component and the reflection component,then the illumination component is adaptively gamma-corrected and contrastenhanced,and finally combined with the reflection component to enhance the image.The result of the single-frame processing is then subjected to improved inter-frame compensation.First,we use CLAHE to process two gray images and use the knowledge of probability statistics to determine whether compensation is needed,after that the brightness difference or noise between consecutive frames will be reduced.The experimental results show that the proposed algorithm is more subjective in overall brightness and more prominent in details.In the objective aspect,the proposed algorithm preserves the structure of the original image better and shows more visual salience,which is more attractive,and less time complexity.