Shadow Detection and Removal for Occluded Object Information Recovery in Urban High-Resolution Panchromatic Satellite Images
Nan Su,Ye Zhang,Shu Tian,Yiming Yan,Xinyuan Miao
DOI: https://doi.org/10.1109/jstars.2016.2570234
IF: 4.715
2016-01-01
IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Abstract:The existence of shadows in very high-resolution panchromatic satellite images can occlude some objects to cause the reduction or loss of their information, particularly in urban scenes. To recover the occluded information of objects, shadow removal is a significant processing procedure for the image interpretation and application. In this paper, we propose a novel framework of shadow detection and removal for panchromatic satellite images to restore the obscured object information. In shadow detection, we present an automatic soft shadow detection method by the combined application of a bimodal histogram splitting method and image matting technique. Soft detection results can exhibit both umbra areas and penumbra areas to describe the shadow distribution precisely. In shadow removal, we propose a spatial adaptive nonlocal sparse shadow removal method to operate at two levels. For the initial step, we apply the line correction method to enhance shadow areas roughly in global. In the refined process, we study the characteristics of objects and shadows, and analyze spatial relationship between them. The second linear radiometric correction and nonlocal sparse model are used to simultaneously control the brightness and smoothness of the recovered shadow areas to be the same as the corresponding nonshadow areas based on group matrix with similar patches. Our method can restore the uniform objects in the shadowareas. High-resolution panchromatic shadow images of different cases and different satellites are processed by the proposed method. The experimental results verify the effectiveness and superiority of the proposed method by comparing three other methods.