An Adaptive Filter to Reduce SAR Speckles

A Case Study of Mapping Inundation Extent on the North Carolina Floodplain Using the JERS-1 SAR Data

Abstract

An adaptive filter was developed to reduce speckles and to preserve the boundaries (e.g., a flooded/nonflooded boundary) on a synthetic aperture radar (SAR) image, and hence to improve the accuracy on the inundation extent mapping using the SAR data. Based on the counts of pixels in each category within a moving kernel, the filter used different filtering approaches to reduce the speckles and retain the boundaries. As an example, the authors used the original, median-filtered, and adaptive-filtered Japanese Earth Resource Satellite – 1 (JERS-1) SAR data to map a flood extent on the North Carolina coastal floodplain, and to investigate the effectiveness of the adaptive filtering through a comparison study of the derived flood extents. Spatial correlation analysis and accuracy evaluation indicated that the adaptively filtered SAR data achieved higher accuracy on the inundation maps than either the median-filtered or the original SAR data.