Performance Prediction in Radar Image Filtering and Lossy Compression. Part II: Compression

This paper analyzes approaches and recently proposed methodology of performance prediction for lossy compression of radar images assuming that they are corrupted by speckle noise. Techniques based on discrete cosine transform in blocks are considered. Characteristics of the speckle are assumed known in advance or pre-estimated with appropriate accuracy. These characteristics are taken into account in setting parameters of compression methods such as quantization step. It is demonstrated that simple statistics of DCT coefficients in 8x8 pixel blocks can be employed for predicting improvement or reduction of peak signal-to-noise ratio and compression ratio. Prediction dependences are obtained using scatter-plot and regression in off-line mode. The proposed prediction approaches are tested for simulated and real-life images.