Multiscale Filtering of Correlated Speckle Using Scale and Space Consistency

ABSTRACT Speckle noise is often correlated due to image post-processing or imperfect imaging transfer function. in order to filter correlated speckle noise in a multiscale framework (stationary wavelet decomposition), we propose two approaches, respectively parametric and non parameteric. The first approach is based on a parametric modelisation of the speckle autocorrelation function using a combination of moving average gamma distributed signals. Then, we derive explicit relations that relate high order wavelet coefficient moments with speckle noise second-order statistics. Numerical results show that wavelet statistics are highly sensitive to speckle second-order properties. The second approach uses a non parametric estimation of the noise energy: at each scale level, the significant wavelet coefficients are modeled probabilistically, and a shrinkage function is derived based on the model obtained. Joint use of space and scale consistency is applied for better preservation of edges and smoothing of homogeneous regions. Mutliscale shrinkage functions are then merged producing an edge map and a filtered image. We show results of experiments carried out on both synthetic (simulated) images as well as on real SAR images. Results indicate that the proposed methods achieve an improved trade-off between speckle reduction and details preservation over conventional speckle filters