Finding backscatter variability where sonar or radar image data are affected by speckle or fading

What is it about?

The speckle (otherwise called fading) in backscatter image data from sidescan sonars and side-looking airborne and other forms of imaging radar is due to the narrow frequency bands of these systems and interference of wavelets backscattered from features on the imaged surface which have differing range. If no other causes of signal variation are present, this is expected to lead to the variability having a Rayleigh distribution (probability density function or PDF). In order to study variations in backscattered intensity, we usually overcome the variability by first filtering the data spatially, but this degrades the effective spatial resolution of the data. An alternative approach is outlined here, after noting that when we convert data to decibels (taking the logarithm of the signal intensity), the effect of the speckle becomes additive rather than multiplicative. Where the underlying backscatter strength of the terrain varies, the recorded signal equals a form of convolution equation containing a PDF representing the varied backscatter strength and a PDF representing the effect of speckle. The alternative approach to resolving backscatter variability therefore involves deconvolving histograms of the backscatter data using the known effect of speckle. This approach was illustrated using sonar data crossing the edge of a submarine lava flow.

Featured Image

Why is it important?

The approach allows the backscatter variability to be resolved in situations where patches of different backscatter are small so the alternative approach of filtering the data would effectively destroy the information on variability wanted.