Shadow Detection and Compensation in High Resolution Satellite Image Based on Retinex
Shugen Wang, Yue Wang · 2009
In this paper, shadow detection and compensation are treated as image enhancement tasks. The principal components analysis (PCA) and luminance based multi-scale Retinex (LMSR) algorithm are explored to detect and compensate shadow in high resolution satellite image. PCA provides orthogonally channels, thus allow the color to remain stable despite the modification of luminance. Firstly, the PCA transform is used to obtain the luminance channel, which enables us to detect shadow regions using histogram threshold technique. After detection, the LMSR technique is used to enhance the image only in luminance channel to compensate for shadows. Then the enhanced image is obtained by inverse transform of PCA. The final shadow compensation image is obtained by comparison of the original image, the enhanced image and the shadow detection image. Experiment results show the effectiveness of the proposed method.