BLOCK ADAPTIVE COMPRESSED SENSING OF SAR IMAGES BASED ON STATISTICAL CHARACTER

Block-based processing has shown promise to reduce computation complexity and storage space for image Compressed Sensing. In this paper, a new architecture for SAR images is proposed, as an improvement for traditional Block Compressed Sensing of natural images. The proposed scheme adopts the basic structure of existing Block Compressed Sensing, and studies the character of SAR images. Based on the difference of statistical property among sub blocks, the proposed scheme can adaptively select the number of measurements that needed to take for every sub blocks. Different from equality measurement, adaptive sampling can sufficiently capture the diversity between sub blocks and keep their properties well. Several numeral experiments also demonstrate that the proposed approach outperforms the existing scheme, achieving comparable reconstruction quality via fewer measurements.