Model-based coding of SAR and ultrasound images

Compression of natural images has received wide-spread attention in the recent decades. Significantly less attention has been paid to compressing radar and acoustic images, such as synthetic aperture radar (SAR) and still less ultrasound. Radar and acoustic images of this type have inherently different statistical properties and thus do not lend themselves to efficient compression using algorithms developed primarily for natural images. The work reported in this paper addresses an approach for handling images with such statistical properties. The approach is based on a image-structure/speckle-texture decomposition. The resulting components are compressed separately using subband coding and texture modeling. In reconstruction, the components are decoded and recombined. The new technique is targeted for low bit rate applications and is able to preserve the perceptually important characteristics, leading to high quality reconstruction