Classification performance of reduced-resolution hyperspectral imagery from JPEG 2000 decompressed data

Oscar Antonio Perez, University of Texas at El Paso


Instruments have been developed over the last few years to acquire Hyperspectral Images (HSI) from the Earth using airborne and satellite sensors. This thesis analyzes HSI from a satellite using the Hyperion Imaging Spectrometer. The spectrometer generates HSI which have a higher bandwidth than a regular image based only on the combination of three primary colors (Red, Green and Blue) since HSI have more than 3 primary components. For any material, the amount of solar radiation that it absorbs, or reflects varies with wavelength. This property makes possible an accurate remote sensing of any given material. ^ The size of digital data produced by the spectrometer is huge which makes the process of an accurate analysis on this data very costly in the following aspects: space for it in storage and the time for its analysis. This thesis proposes the losslessly usage of the JPEG 2000 standard by using the resolution scalability tool that it provides. By using this tool the amounts of data, time for analysis, and time for transfer could by reduced while a high hit rate percentage could be achieved. ^ The main objective is to reduce the resolution of the original dataset while maintaining the quality of the data to be able to obtain excellent results on remote sensing of aluminum-rich and aluminum-poor mica. As stated in the following chapters a two step process is needed to acquire the maximum hit rate. The data was reduced by one through five levels of wavelet decomposition. Each of them yielded a new dataset which was analyzed to check if they met the goal of reducing the cost on the aspects previously listed. ^

Subject Area

Engineering, Electronics and Electrical|Remote Sensing|Computer Science

Recommended Citation

Perez, Oscar Antonio, "Classification performance of reduced-resolution hyperspectral imagery from JPEG 2000 decompressed data" (2004). ETD Collection for University of Texas, El Paso. AAIEP10803.