The growth in the use and storage of multi and hyperspectral image data sets has necessitated the use of image compression for telecommunications, storage, and dissemination of this imagery due to the sheer volume of data. The need for high-quality low bitrate compression cannot be met with the current technological standard, JPEG. New technologies have been developed which significantly outperform JPEG. Wavelet/TCQ is among the best compression algorithms in the scientific literature and has been selected by the ISO JPEG standards committee as the baseline algorithm for the newly developing JPEG 2000 standard. In Phase 1, we will develop a quad-tree spatial segmentation technique based on maximizing coding gain to add to JPEG-2000. Several different linear decorrelating transforms will be examined including band prediction and the KLT and will be compared against a 3D wavelet transform. Additionally, spectral segmentation of multispectral data will be developed based on global band correlation. Phase II will involve the investigation of more sophisticated segmentation techniques and design of a hardware/software system for the compression multispectral and multicomponent imagery. Numerous commercial and governmental applications utilize multispectral imagery. Applications include remote sensing, reconnaissance, and mapping. Medical imaging, telemedicine, and medical databases can also benefit from this technology. MRI and CAT data sets share many similarities with multispectral imagery.
