Drastic improvement of change detection results with multilook complex SAR images approach

Abstract

Coherent Change Detection (CCD) is a powerful technique that uses Synthetic Aperture Radar (SAR) coherence to measure subtle ground changes in the imaged area. Unfortunately, the coherence estimator is biased for low coherence values, resulting in a highly degraded change detection performance. The spatial multilooking technique is typically used to improve coherence estimation but at the expense of spatial resolution. Actually, there are some SAR satellites that are able to deliver Multiple Look Complex (MLC) SAR images, which provide noticeable coherence bias reduction. In the present work, we investigate detection performance improvement that can be obtained through the use of MLC SAR images. The detection probability and false alarm are evaluated using experimental very high-resolution SAR data. After SAR image focusing and coherence estimation, the results indicate that the use of MLC SAR images with four looks allows for nearly 60% higher detection probability in the case of a low false alarm rate.