An Intelligent Image Compression Broker Based on Objective with Feedback

In this Phase II proposal, we plan to build an augmented intelligent target compression broker based on our result from our Phase I work. In addition to the already developed modules to handle EO and IR imagery, we will equip the upgraded system with the capability to handle SAR, multi-/hyper- spectral imagery and LiDAR data. SAR image compression techniques such as 3D SPIHT and sparse representation based on techniques will be investigated. A multi-/hyper- spectral image compression method that is also based on sparse representation will be implemented and tested. Two advanced matching pursuit algorithms, JOMP and FOMP, developed to sparsely represent the compressed data will be evaluated. For LiDAR data compression, a novel geometry-based LIDAR compression technique is proposed. It first performs LIDAR classification to classify the data into tree points and building/ground points. Next, different compression methods are proposed to compress the tree cloud points and building/ground cloud points differently. Methods to quantify the quality degradation due to compression will be developed for SAR, multi-/hyper- spectral images as well as LiDAR data. Finally, the proposed system will be validated both subjectively and objectively.