Analysis of Model based Shadow Detection and Removal in Color Images

Abstract

Shadow detection and removal efficiency in color images has become most important requirement while processing aerial images. The successive thresholding scheme (STS) is presented in this paper. Scheme increases shadow detection accuracy. In this paper, we have modified ratio map to obtain accurate gap between shadow and non-shadow pixels. The idea is derived from original Tsai's method. The global thresholding scheme is used to classify the pixels into shadow and non-shadow class. The coarse map obtained for candidate shadow pixels is then processed using connected component and candidate shadow region pixels are grouped. The iterative method is used for detecting true shadow pixels. Experimental results show that, the shadow detection correctness of our suggested model based algorithm is analogous to Tsai's algorithm. Dataset of images is tested using both approaches to calculate precision (P), recall (R) and Fscore (F). The results show outstanding performance for our method.