Effect of Speckle Noise for Detection of Damaged Building Area Using SAR Intensity Images

Abstract

One of the remarkable characteristics of synthetic aperture radar (SAR) is to record physical value called backscattering coefficient of the earth's surface, regardless of weather condition or sun illumination. Therefore, SAR is a powerful tool which can be utilized to develop a universal method to fully comprehend damaged areas in natural hazards such as earthquakes, forest fires and floods. We performed a feasibility study on backscattering characteristics of damaged areas in the 1995 Hyogoken-Nanbu (Kobe), Japan earthquake using the pre- and postevent ERS-1 images, revealing that the backscattering coefficient and intensity correlation between the two attained values were significantly lowered in hard-hit areas. The evaluation, however, was performed without speckle noise reduction. In this study, we investigated the effect of pixel-window size in evaluating building damage using the difference in the backscattering coefficient and correlation coefficient of pre- and postevent ERS images filtered for speckle noise reduction. From the above analysis an optimum window size for the damage evaluation was obtained. It was also found that the accuracy of damage detection was not significantly improved for specklereduction filtering of window size larger than 21×21 pixels.