A new hyperspectral image compression using radon transformation maco optimization and dvat svd techniques

Abstract

Hyper Spectral Image (HSI) compression has recently become a popular newlineresearch area in remote sensing applications. It is a challenging and demanding task, newlinebecause it has large number of spectral data. Optical remote sensing is much increased newlinedue to newly imported sensor technologies and advancements. Moreover, it exhibits newlinesignificant spectral correlation, whose exploitation is crucial for compression. This newlineresearch work proposed different compression techniques for HSIs. In the initial newlinephase of this work, a lossy compression techniques, namely, Residual Dependent newlineArithmetic Coder (RDAC) is designed for HSIs. The main objective of this work is to newlinereduce the complexity while compressing the large volume of data by compressing newlinethe spectral bands. In this module, the Gray Level Co-occurrence Matrix (GLCM) newlinetechnique is employed to extract the texture features of the given HSI. Then, the kmeans newlineclustering technique is utilized to select the reference band in each cluster newlinebased on the cluster prominence value. Furthermore, the RDAC technique is used to newlinecompress the reference band and the residual band information of each cluster. newlineFinally, the compressed image is decompressed to obtain the original HSI. newlineTo improve the clustering efficiency based on the optimal solution, a novel newlineModified Ant Colony Optimization (MACO) is integrated with the RADON newlinetransformation technique in the second phase of this work. Here, the median filtering newlinetechnique is employed to preprocess the given HSI, which efficiently removes the newlinenoise in the image. Then, the single band image is selected from the original HSI and newlinethe color features of that band is extracted by using the HIS model. Then, the newlineproposed MACO technique is applied to select the band index based on the fitness newlinevalue. The multi-thresholding technique clusters the single band image into 6 newlinesegments. After that, the RADON transformation and Zig-Zag encoding techniques newlinevi newlineare applied to compress the HSI band. Finally, the original band image is newlinereconstructed by performing the Zig-Zag d