Research On Shadow Detection And Removal Algorithms Of Static Images

As a ubiquitous natural phenomenon,shadow has always been the focus and difficulty in image processing.In recent years,shadow processing technology has made great progress in the field of computer vision.As an important preprocessing step in the field of computer vision,shadow processing has been widely used in various fields such as target detection,medical images,aerial imaging and so on.However,in recent years,shadow processing research has been more focused on shadow processing in video sequences.Shadow processing algorithms based on video sequences usually use the frame difference method or background subtraction method to extract shadow features,these algorithms are difficult to apply to static images.The research content of this thesis is the shadow processing research of static images.Shadow detection and shadow removal are performed on the images and the feasibility of the algorithm is verified through experiments.The specific work is as follows:(1)In view of the obvious problem of the boundary between the restored area and the illuminated area in the shadow processing of the static image,a shadow processing method for the restoration of the area illumination is proposed.Based on the maximum between-class variance method,a double-threshold Otsu pixel classification method for shadow detection is proposed.In order to make the edge recognition more accurate and the color rendering more natural,Canny edge detection is performed on the shadow area in the gradient domain to achieve accurate identification of the shadow edge.Finally,the shadow pixel coefficient is restored by calculating the average illuminance and intensity ratio of the shadow and non-shadow area pairs.Make the illuminance of the shadow area consistent with the light area,so as to achieve shadow removal.Experiments show that this method has better recovery effect on the shadow boundary.(2)Aiming at the problem of artifacts appearing in penumbra areas during the shadow removal process,a method for eliminating shadow areas with uneven intensity is proposed.This method can retain texture information from a single image marked by a simple user.The shadow mask is detected by analyzing the shadow feature marks specified by the user.In order to avoid artifacts caused by excessive removal,intensity sampling is performed around the shadow boundary to determine the intensity range.Then the texture noise in the samples is eliminated by bilateral filtering,and the shadow scale is estimated by the image pyramid.The error in the estimated shadow scale is eliminated by local smoothing,and finally the inverse Hadamard transform is used to obtain a shadowless image.Apply color correction to eliminate color errors due to image processing.Experimental results show that this method can effectively remove penumbra and preserve texture consistency between regions.