Research On Shadow Removal In Remote Sensing Images Based On Linear Correction

In recent years,with the availability of high spatial-resolution satellites such as QuickBird,IKONOS,GeoEye,Remote sensing technology has an extensive application and has bring a large number of remote sensing images.Remote sensing technology are used in many fields.Such as target detection,target recognition,image segmentation and so on.However,Shadow formed by elevated objects such as high buildings,bridges,and trees in most of remote sensing images.On the one hand,shadows information can be use to 3-D shadow reconstruction,building height estimation,and position recognition.On the other hand,the presence of shadow also causes negative impacts,in order to remove these negative impacts,we must remove the shadow in remote sensing images.Shadow processing mainly included two aspects:shadow detection and shadow removal.In this paper,we will focus on the shadow removal mainly.In this paper,we found that the exiting linear correction shadow removal algorithm proposed may cause the non-homogeneous shadow area with color distortion.In order to resolve this problem,we proposed a adaptive bright and dark classification linear correction algorithm.In this method,we extract the shadow and non-shadow samples,decompose the input image into uniformly patches,then traversal to calculate the contrast and find the largest contrast,calculate the average as the threshold,finally match the shadow and non-shadow and calculate radiation correction parameter,In order to smooth the boundary and improve the quality of compensation image,the method show a mean weighted algorithm.The results show that the method in this article can weaken the image color deviation,smooth the boundary,the transition of shadow boundaries will be more natural.there is still exit color shift and boundary texture lost for the adaptive brightness classification compensation algorithm,This article provided a improved algorithm that based on guided image filter with linear correction.firstly,we extract the shadow and non-shadow as matching data,judgment the image shadow class k and Use the maximum between-class variance(OTSU)to classify the shadow area.Let k=k-1 Each classification,when k equal to 1 to stop classify,then the shadow and non-shadow regions are matched between class and class.Calculating their correction parameters,in order to smooth the boundary trace and save the boundary texture details,the theory combined with the guided image filter and linear correction algorithm,The results show that the method in this article can enhance color information,smooth the boundary trace,and maintain the boundary texture details.