ISAR Image Segmentation by Non Linear Diffusion Equation

This paper describes a non linear spatial filtering for the target radar segmentation of ISAR/SAR images. Since such images are usually corrupted by speckle noise and clutters, image processing technique are required in order to obtain a more clear image. We have implemented a spatial technique based on the non linear diffusion equation, also known as the Perona-Malik equation. Our model is based on an calculation of the spatial gradient on a 5 X 5 stencil in order to increase the merging of target corresponding regions. By suitably estimating a diffusion coefficient related parameter, the gradient threshold, target segmentation is done into few integration steps of the equivalent explicit scheme. The resulting images maintain most of the original details also visible in converting the filtered image in a binary format. The proposed method can be the first step of target classification by matching correlation of more advanced techniques.