Adaptive Change Detection With Significance Test

In this paper, we propose a significance test-based change detection method that can automatically discriminate between changed and unchanged pixels in the difference image. The method adaptively considers the local contextual information, which is contained in the neighborhoods of each pixel, to derive the decision threshold. In our method, a significance test algorithm based on maximuming a posteriori estimate is constructed; then, a weight to each pixel in the block is imposed to increase the change detection accuracy. In our proposed method, the distribution of the difference image satisfying Laplace model also leads to good precision. For the experimental component, two types of images were tested. And experimental results proved the effectiveness of the significance test method when compared with four state-of-the-art change detection methods.