SAR Image Formation by Pixel Classification (IF-PC)

Abstract

This paper presents a novel synthetic aperture radar (SAR) image formation technique—SAR Image Formation by Pixel Classification (IF-PC). The technique is integrated with the standard backprojection image formation algorithm to form SAR images with significant reduction in side lobes and noise. The key innovative concept in the IF-PC technique is the formation of many realizations of a SAR image of the same scene from many instances of randomized and reduced apertures. The IF-PC algorithm classifies each pixel in the SAR image into either target class (associated with physical objects) or noise class (noise, sidelobes) based on the statistics of the image pixel from many realizations of the SAR image. The IF-PC technique generates both types of resulting SAR imagery: amplitude and complex images that are almost noise-free. This paper demonstrates the results using both simulation and real radar data from the U.S. Army Research Laboratory's (ARL) low-frequency SAR radar in forward-looking mode.