Speckle reduction of SAR imagery using homomorphic processing and predictive filtering

This paper describes a method of reducing the speckle in a SAR image using a variation of homomorphic processing. In a SAR image, speckle is typically characterized as multiplicative noise such that traditional means of additive noise reduction may not perform well. In homomorphic processing, a signal is first passed through a logarithmic function, which would transform any multiplicative components (such as speckle noise) into additive components in the logarithmic domain. At this point, any additive noise reduction technique may be used for denoising. Here, we incorporate predictive filtering into the homomorphic process as a means to reduce the speckle.