Keywords : speckle, filtering

Speckle Filtering

Speckle is a particular noise present in all sonar images. It has for origin a summation of various echoes from point-targets which are located at slightly different distances from the sonar. The range variation is the same as the signal wavelength, as the transmit wave is "monochromatic" the echoes summation of the ensonified area is done in a more or less coherent way, from a sonar sample to another one can go from a constructive summation (signals in phase) to a destructive one (signals in phase opposition). From a statistic point of view, the resulting amplitude follows a Rayleigh law which is translated on the image by this granular aspect named "speckle"

Speckle is a multiplicative noise because the echo summations are independent from the seafloor reflectivity, they only depend on the micro-relief. Classical filtering techniques which are used on additive noises cannot be directly used on sonar image speckle. Many authors have brought contributions to improve the speckle filtering techniques, mainly on radardata. Filtering is aimed at removing the noise without loosing the image resolution. Within the framework of a collaboration between Ifremer and the Timisoara Polytechnique University (Romania), researches are conducted by Professor Alexandru Isar's team on sonar images speckle filtering. The investigated techniques are based on wavelet decomposition s. The following images show the latest results obtained by Pr. Isar.

Swansea wreck. Klein 5500 sonar. Courtesy of DGA/GESMA Images in Black & White

Swansea wreck. Klein 5500 sonar. Courtesy of DGA/GESMA - Coloured images

Zoom on ribs of wreck

This image features several "high-frequency" objects which are degraded by classical filters.

"High-frequency" objects are preserved.

Raw sonar image in N&B and in colour

Filtered images in N&B and in colour