Comparative Analysis of Single Image Shadow Detection and Removal in Aerial Images

Abstract

The article for the specific region segmentation is discussed here. This paper compares the work done on the contour model based shadow detection algorithm with the random field model based shadow detection method. The contouring based shadow detection model adapts the traditional geometric active contours that the contour is steadily subjective toward the shaded region and the shadowy regions in the picture. The shadow region segmentation follows the post processing for optimal threshold calculation. The complexity metric is constructed to separate the boundary of true shadow regions. Whereas the random field model based on shadow exposure process uses a superficial outside descriptor, and colour shade area, to describe the colour surface variants. The conditional random field is combined with the colour descriptors to locate lighting pair and acquire the shaded portion. The shadow region riddance by neighbourhood shade constancy utilizes anisotropic dispersion to guess the local image pixel illumination in shadow region and provides the better performance in shadow region illumination enhancement.