Publication: Complexity reduction in integer klt for lossless hyperspectral image compression
Loading...
Date
2022-08-01
Authors
Rajan, Dhurgesh Nair
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
The research presented in this thesis is concerned about complexity reduction in Integer KLT algorithm for lossless hyperspectral image compression. The Integer KLT is addressed because it outperforms other algorithms in decorrelating the spectral component in hyperspectral images. At the same it has it’s own drawbacks. Integer KLT has many process which makes the algorithm to be more complex and time consuming. The goals of this study are to reduce the execution time and complexity of the Integer KLT algorithm by modifying the code, as well as to implement and compare the performance of the Integer KLT algorithm on the Raspberry Pi 4 and the desktop. Two method were used to reduce the execution time which are simplifying the mathematical expression and removing tiling method. The execution time for the compression of the AVIRIS and Hyperion hyperspectral image datasets improve with the following method. Next, a different platform was used to implement the Integer KLT algorithm to evaluate the performance. From the following method, Desktop’s Windows shows a superior performance when compared to Raspberry Pi 4. This result was influenced by number of cores and RAM respectively of each hardware. The window’s desktop has a high power system with a processor of eight cores and 16 logical processors compared raspberry pi 4 where it only has four cores. This research justifies that an hardware specification and the complexity of coding language, highly influenced the execution time of a program.