Research On Shadow Detection And Compensation In Remote Sensing Images

With the increasingly widespread application of remote sensing technology, remote sensing images start to be used widely and widely in many areas in recent years. However as the resolution of the satellite images increasing, shadows cast by cultural features from the images take on special significance. On the one hand, they cause the partial or total loss of information in the shadow areas, and consequently they make image-analysis decline. On the other hand, we may take advantage of the presence of shadows as a valuable cue for height estimation. In order to improve the image quality and effectively apply the images, shadow processing technique in RS images become is a hot research problem. However, facing the mass RS images, how to detect and compensate the shadow in RS images is also a tough problem. Therefore, it’s indeed extremely meaningful to research shadow detection and shadow compensation. The shadow processing technique includes two parts which are shadow detection and shadow compensation.This paper analyses the shadow detection approach based on the c3component and improves this method. Firstly, in order to minimize the noise effect, the c3image is smoothed. Then, extract shadow boundaries by means of Sobel operator. Finally, in order to separate water from shadow region extracted, this paper examines the variance of groups of pixels. This paper presents a procedure based on c3component and Sobel operator. The experiment shows that this method can detect shadows quickly and accurately and can separate water from shadow region effectively. Therefore, this procedure enhances the stability of the method based on c3component, and increases the robust.This paper analyses the suppressing B component method and Linear-Correlation Method, and applies the shadow feature that shadow region is matched to their companion region. After improving and merging these two methods, this paper presents a shadow compensation method based RGB and HSI color space. Experimental results indicate that this proposed method can improve the visibility of features in shadowed areas and retain the value of non-shadowed regions. The results show the properties and applicability of the proposed approach is good.In this paper, some experimental parameters are obtained by the manual selection, without effective theoretical basis, the method of parameter settings and the improvement and application expansion of shadow processing technique will be researched in the future.