Paper Title
Shadow Detection and Removal for Occluded Object Information Recovery in Urban High-Resolution Satellite Images using Exemplar based Image in Painting

High resolution satellite images can be used for various applications such as producing composite images of an entire hemisphere, to map a small area of the earth, meteorology, oceanography, fishing, agriculture, forestry, education etc. But the presence of shadows in high-resolution satellite image can occlude some of the objects in the image. This problem mainly caused in urban areas. Shadows are usually cast by long objects such as buildings and towers in urban areas. The existence of shadows may cause loss of feature information, false colour tone and shape distortion of objects, which seriously affect the quality of images. In order to restore the obscured object information shadow detection and shadow removal is an essential preprocessing step. This paper propose a novel framework of shadow detection and removal for high-resolution satellite images. For detecting the shadow, initially a binary mask image is constructed using hard thresholding. Using morphological operators a final mask image is obtained, which distinguishes shadow and nonshadow area precisely. Finally an image matting technique is used to solve the presence of penumbras based on the final mask image. For shadow removal, this paper proposes a robust exemplar-based image in painting algorithm. This methods iteratively search the source region or the nonshadow region and fill the missing region i.e., shadow region, with the most similar patch in the nonshadow region. The spatial information in the nonshadow region is utilized by using a segmentation method which improve the performance of in painting. The proposed method automatically selects parameter values of the robust priority function, adaptively determines patch size, and reduces search region with the segmentation map. As a result, a shadow-free satellite image is obtained with same quality across the shadow and nonshadow area. The experimental results verify the effectiveness and accuracy of the proposed method by comparing with other methods. Keywords - Hard Thresholding, Image Matting Technique, Image Inpainting, Graph Based Segmentation.