Two adaptive filters for speckle reduction in SAR images by using the variance ratio

Abstract

This paper presents some experimental results on the properties of image speckle along with two adaptive filters that are useful for speckle reduction. An investigation of available SIR-B digital image data over Australia shows that speckle is non-white Gaussian noise and fits a multiplicative noise model in which the noise, uncorrelated with signal, has a mean of I and a constant standard deviation. The non-uniform spectrum can be represented by an empirical formula. Based on these results, two adaptive filters have been designed, a moving average filterand a combined moving median filter. In both of these, variance ratios are used to govern the filter parameters. The filters have been applied to some SIR-B image segments, showing good results in both smoothing speckle and maintaining edges.

An automated region growing algorithm for segmentation of texture regions in SAR images

“…Considering this, the quantization level is reduced to 64 since no signi® cant loss of information is observed, resulting in a manageable co-occurrence matrix with less computational load. In addition to this, since the correlation between neighbouring pixels falls o rapidly (Li 1988 ), the magnitude of b is taken to consider only eight connected pixels.…”

A New Method for SAR Speckle Noise Reduction (CST Filter)

“…The important feature of Li's method is the change in the size of the filtering window based on "the variance ratio" (Li, 1988). The variance ratio is defined as the ratio between the variances of the noise-free image and the observed image in the filtering window.…”

“…In this paper we selected five various spatial filtering methods that use local statistics. The reviewed methods are as follows: (Refined) Lee filter (Lee, 1981); Sigma filter (Lee, 1983(Lee, , 1986); Weighting filter (Martin and Turner, 1993); Us method (Li, 1988);…”

A model for removal of speckle noise in SAR images (ALOS PALSAR)

“…The Wiener filter (Walkup and Choens, 1974) and other filters with criteria of minimum mean-square error (MMSE) are examples of filtering algorithms that are based on the frequency domain (Li, 1988). The wavelet approaches have been used to reduce speckle in SAR images, following the theoretical basis for multiresolution analysis of Mallat (1989a;1989b).…”