SAR Image Compression with Vector Quantization of Wavelet Trees at Low Bit Rates

A novel wavelet-based vector quantization (WVQ) is presented for SAR image compression at low bit rates. Through analysis of the decomposed wavelet coefficients' statistic property, establish directional vectors based on spatial orientation tree (SOT) structure and apply LBG algorithm to train codebook and vector quantization of wavelet coefficients. For a typical SAR image, the reconstructed images coded by WVQ achieve gains 0.1 dB to 1.0 dB on average in PSNR and preserve superior perceptual quality compared with the coding results of set partitioning in hierarchical trees (SPIHT) algorithm at low bit rates.