Optimized rate allocation and background classification applied to compressed hyperspectral images using JPEG2000 Part 2
In this thesis we show how to use JPEG2000 for compressing multi dimensional data. Here we specifically employed a technique for compression that is more tolerable for the mean squared error (MSE). For the case study we used hyperspectral data from the Hyperion, because of its highly correlated nature which could be exploited for data compression. One of the objectives of this thesis is not only to measure the compression performance of the Discrete Wavelet Transform (DWT), but also to evaluate the performance of the data when using lossy algorithms. We first applied the transform in the spatial direction and compressed it band by band using JJ2000. Since, JPEG2000 Part 2 does not specify a method for bit allocation we incorporate the rate distortion optimal (RDO) as the bit allocation for each of these bands. Here we have employed the mixed model to find the rate distortion curves and then incorporated these with the RDO, and compared our results to the traditional high bit rate quantizer model. It has been observed that this approach gives lower MSE than the traditional but these results don't concur with the background classification. The traditional model does better in the case of background classification.
Attluri, Silpa, "Optimized rate allocation and background classification applied to compressed hyperspectral images using JPEG2000 Part 2" (2005). ETD Collection for University of Texas, El Paso. AAI1430215.