| Speckle is a common phenomenon in all types of coherent energy imaging systems such as Synthetic Aperture Radar (SAR), laser, ultrasound, acoustics, sonar etc. Speckle is a multiplicative-convolutional noise and as such is different from other commonly found types of noise such as Additive White Gaussian Noise (AWGN). Hence different methods of processing are required to restore speckled images. Moreover, in many applications, the edge structure of an image is very important, and usual filtering methods are not well suited for preserving edges particularly in speckled images. In this work, an extensive study has been made to investigate the applicability of different existing nonlinear filtering methods and also a new Quadratic Volterra Filter (QVF) based on speckle-model to solve the problem of speckled image restoration in terms of noise smoothing and edge preservation. Edge detection itself on speckled images is a major problem which has not been addressed by many researchers. This thesis attempts to provide a better approach to the solution of restoring images corrupted by speckle while preserving their edge information. |
