Change Detection in Multispectral Satellite Images using Histogram based Thresholding Technique

Abstract

In this paper, an unsupervised global threshold technique, which is based on histogram analysis, is proposed. This technique analyzes the histogram. The histogram is partitioned in intensities at equal interval, and each interval assigned as single intensity. Then, differentiation is applied to get location of maxima in histogram. Depending on the locations of maxima, threshold is calculated to separate the two regions as change and no-change. The utilization of the differentiation improves the visualization of the maxima of different regions. To demonstrate the effectiveness of technique, experiments are conducted on two multispectral satellite images.