Speckle reduction by spatial filtering

Synthetic aperture radar (SAR) is a technique that allows improved image resolution through the coherent processing of a sequence of radar echoes. The coherent signal received after reflection is processed so as to synthesize a long antenna in the direction of platform motion. However, images obtained by a coherent sensor (such as a SAR or laser) are contaminated by a kind of noise called speckle, which comes from multi-path interference of coherent waves scattered from a distributed target. Due to this type of noise, the image of a homogeneous surface with constant backscatter coefficient will show brightness variations. The probability density function (PDF) of the speckle intensity (under the assumption of a large number of statistically independent scatterers, with the phase uniformly distributed between 0 and 2 π) is an exponential distribution, with standard deviation equal to its mean, and with the most probable intensity value at any point in the image equal to zero. To combat speckle noise there are effectively two approaches. The first introduces incoherency through the sensor itself, and the second processes the speckled image using spatial or temporal filtering techniques. In this paper, the authors deal essentially with the first of these approaches