Despeckle-based SAR image compression

SAR images suffer from speckle noise that degrades image quality. Speckle removal is to he applied when signatures such as edges and fields have to be extracted from the image; speckle removal is also needed for image compression as such noise increases image entropy. Nevertheless, most lossy compression skim act as low pass filters that do not fit multiplicative noise reduction. In order to build a compression skim that takes into consideration statistics of SAR images, a well known reflectivity estimator is integrated into an image compression strategy, so that image compression artifacts may be viewed as an adaptive despeckled filter instead of a simple low-pass filter. The compression method has been developed with the most efficient tools dedicated for image compression, that suit the JPEG 2000 norm, such as multi-resolution analysis achieved by wavelet transform, optimal scalar quantization adjusted with a distortion-rate constraint and an arithmetic encoder. It yields a compression algorithm that acts in a similar way to the Lee filter. A theoretical analysis is also applied on the multi-wavelet transform in order to take into consideration scene correlation in the speckle removal filter. Some results are presented on ERS PRI images compressed at a factor of η = 20 : 1