Effect of lossy vector quantization hyperspectral data compression on retrieval of red-edge indices

This paper evaluates lossy vector quantization-based hyperspectral data com pression algorithms, using red-edge indices as end-products. Three compact airborne spectrographic imager (CASI) data sets and one airborne visible/in frared imaging spectrometer (AVIRIS) data set from vegetated areas were tes ted. A basic compression system for hyperspectral data called the "referenc e" system, and three-speed improved compression systems called systems 1, 2 , and 3, respectively, were examined. Five red-edge products representing t he near infrared (NIR) reflectance shoulder (Vog 1), the NIR reflectance ma ximum (Red_rs), the difference between the reflectance maximum and the mini mum (Red_rd), the wavelength of the reflectance maximum (Red_lo), and the w avelength of the point of inflection of the NIR vegetation reflectance curv e (Red_lp) were retrieved from each original data set and from their decomp ressed data sets. The experiments show that the reference system induces th e smallest product errors of the four compression systems. System 1 and 2 p erform fairly closely to the reference system. They are the recommended com pression systems since they compress a data set hundreds of times faster th an the reference system. System 3 performs similarly to the reference syste m at high compression ratios. Product errors increase with the increase of compression ratio. The overall product errors are dominated by Vog 1, Red_r s, and Red_rd, since the amplitude of product error for these products is o ver one order of magnitude greater than those for the Red_lo and Red_lp pro ducts. The difference etween the overall error from the reference and that from system I or 2 is below 0.5% at all compression ratios. The overall pro duct error induced by system I or 2 is below 3.0% and 2.0% for CASI and AVI RIS data sets, respectively, when the compression ratio is 100 and below. S patial patterns of the product errors were examined for the AVIRIS data set . For all products, the errors are uniformly distributed in vegetated areas . Errors are relatively high in nonvegetated and mixed-pixel areas.