Research On Color Deviation Detection And Correction Based On Color-feature Analysis

In recent years,with the rising popularity of mobile Internet and smart phones,the way of image acquisition is gradually increasing and image data grows significantly.Due to the imaging scene condition,images taken in unknown imaging system or light source have color cast issue.Using image processing methods like color cast detection and correction,we can effectively remove the color deviation of the image so as to improve the overall visual quality of the image.This paper only considers the color shift which is caused by the changes of the environmental illumination,and the main research is to detect and remove a color cast in uniform light source.A cast detector,based on a certain principle,detects whether the image has a color cast and classifies the input images.A cast remover is then applied in cases of evident cast.Based on the two-dimensional histograms algorithm,we will mainly study how to eliminate the influence of dominant colors and classify the input images as presenting no cast,evident cast,a predominant color in this paper.We analyze the distribution characteristics of image colorfulness and color naturalness,and propose a new cast detection approach based on color-feature analysis.Applying the distribution characteristics of two-dimensional histograms and image color-feature to cast factor,we can realize the cast detection and classification correctly.Based on the Grey World algorithm,we will mainly study how to overcome the defects of it for images with less color types or Large monochrome blocks.We analyze the relationship between the image colorfulness and color information,and propose a new cast correction approach based on colorfulness weighting.By colorfulness weighting,We discard the image blocks which have simple color,and make the mean of remaining image blocks as the image's global mean,to constraint the gain coefficient and prevent the performance degradation of Grey World algorithm for images with less color types,and improve the scope of the Gray World algorithm.