Research On Shadow Detection And Removal Approach For Single Images

As an unavoidable light phenomenon in real scenes,shadows bring troubles to the comprehension of vision information,as well as increase the error in target recognition,image segmentation,scene analysis and tracking process,and seriously affect the precise analysis and processing of image or video.This paper addresses the shadow detection and removal approaches from single images,and mainly focuses on the method based on 3D intensity surface modeling and shadow feature.The paper first introduces the background,significance and current research status of shadow detection and removal form signal images.And then,this paper summarizes the widely-used methods of shadow detection and removal in recent years,addresses the development trend,as well as the advantages and disadvantages of different algorithms.Finally,this paper briefly describes the imaging principle,classification and characteristics of shadow,as well as the challenge problems in shadow removal.This paper presents a shadow removal method based on 3D intensity surface modeling.Firstly,the method divides the image pixels into shadow and non-shadow ones using support vector machine(SVM),and then adds the shadow and non-shadow pixels to the shadow and the non-shadow region respectively using iterative region growth.By this way,the shadow region can be detected.As for the shadow removal,the paper achieves this process based on 3D surface modeling.First,obtain the smooth cartoon image using image decomposition technology.Second,model the intensity surfaces in shadow region(modeled region)based on that in non-shadow region(modeling regions)with the same texture.Finally,the shadow is removed from signal image using intensity compensation.Experimental results demonstrate that perfect effect can be achieved using the proposed method.At last,the paper presents a shadow removal algorithm based on block-matched adaptive light transfer.The method detects a shadow region using a user-assisted method,and divides the input image into overlapping blocks using image decomposition technique.For each block in shadow region,the method searches for its most similar light region block in non-shadow region to obtain a pair of light-shadow blocks.By processing the light-shadow blocks with the adaptive light transfer method to calculate the pixel values of shadow block,and then remove shadow through coherent optimization processing.The Experimental results show that this method can effectively remove shadows and well recover the texture detail in shadow region.