Superparsing based change detection in high resolution remote sensing imagery

Abstract

In this paper, we present a method to detect changes in high resolution remote sensing images based on superparsing proposed by Tighe et al. By comparing with several superpixel segmentation methods, we choose the SLIC (Simple Linear Iterative Clustering) method which can keep image boundary, produce consistent superpixels with similar size and shape, and also calculates fast. After superpixel segmentation, we obtain the category of each pixel in remote sensing images by using superparsing, therefore we can find change areas easily by comparing their category labels directly. Experiments on two Geo-Eye1 high-resolution remote sensing images demonstrate the effectiveness of our proposed change detection method.