Transform coding of synthetic aperture radar (SAR) images

Transform Coding is a type of source coding that is used to compress the data required for images without substantial loss of fidelity. Transform Coding has been studied by many authors for optical images and the results have shown favorable compression ratios on the order of 10 to 1 without noticeable loss of picture quality. This is due to the highly correlated nature of typical images. Observable autocorrelation lags for SAR imagery seldom occur past lags of 3 or 4. However, since correlation exists, a reduction of the data can be achieved. The approach taken in this paper was to achieve data compression on SAR images by the use of Karhunen-Loeve, Walsh and Cosine Transform coding. All coefficients of each transform were allocated a variable number of bits using the Huang and Schultheiss [1] approach followed by the Lloyd-Max quantizer [2,3]. Experiments were done with several variations of the basic transform coding scheme.