Nearly Lossless Compression of Multispectral Images Using Band Classification Based on Karhunen-Loeve Transformation and Integer Wavelet Transformation

摘要

In this paper, an algorithm using band classification for nearly lossless compression is proposed. The algorithm is based on Integer Wavelet transformation (IWT) as described in [1] and Karhunen-Loeve transformation (KLT) as described in [2] but unlike the describe in the paper [2]. The Karhunen-Loeve transformation has been varied by using band classification, in order to obtain a better approximation and speed of the orthogonal transformation. The algorithm of the Karhunen-Loeve transformation / integer wavelet transformation / EZW implements the compression of the 3D multispectral image data, applying KLT to remove the spatial redundancy across the spatial dimension and using IWT for decorrelation along the spectral dimension. The results show that the BC-KLT/IWT/EZW (Band Classification KLT/IWT/EZW) exhibits a better performance than the traditional algorithm of KLT/IWT/EZW.