Wavelet based multi-dimensional image coding algorithms

A new embedded, block-based, wavelet transform coding algorithm of low complexity is developed for hyperspectral image compression: Three-Dimensional Set Partitioned Embedded bloCK (3D-SPECK). 3D-SPECK efficiently encodes 3D volumetric hyperspectral image data by exploiting the dependencies in all dimensions. It can generate either Signal-to-Noise Ratio (SNR) scalable or resolution scalable embedded bitstreams. It can also generate Region-Of-Interest (ROI) retrievable bitstreams.; To obtain an SNR scalable bitstream, the same indexed bit planes need to be put together across blocks and subbands so that the higher bit planes precede the lower ones. SNR scalable 3D-SPECK follows the subband order from lowest ones to highest ones to search significant coefficients. It maintains the same significance threshold across subbands until exhausts them.; To generate a resolution scalable bitstream, bits in blocks belonging to subbands of coarser scale are put together before those of finer scale. Resolution scalable 3D-SPECK encodes each subband separately to generate SNR scalable sub bitstreams. It maintains separate link lists for each subband and proceeds from the highest to lowest threshold in every subband before moving to the next one.; Applying resolution scalable 3D-SPECK on each tree separately, we can generate an ROI retrievable bitstream. The user can therefore select any ROI and decode that ROI only.; We show that 3D-SPECK, a wavelet domain algorithm, like other traditional time domain hyperspectral image compression algorithms, can preserve important spectral profiles well. We also include two well-known remote sensing classification methods to test 3D-SPECK compressed image sequences. It is demonstrated that the distortions in the reconstructed data caused by the compression process result in only minor losses in classification accuracy even at very low bit rates.