Hyperspectral image compression through spectral clustering
Kais Siala, Amel Benazza‐Benyahia · 2004
In this paper, we are interested in coding exactly and gradually hyperspectral image data. To this purpose, vector lifting schemes (VLS) are retained since they take into account the spatial and spectral redundancies in a multiresolution way. However, the high value of the number of components (some hundreds) prevents us applying directly the VLS, due to the tremendous operational complexity. Our contribution consists of a specific preprocessing of the hyperspectral images to make possible the use of VLS at the further stage. Experiments performed on AVIRIS images indicate the outperformance of the proposed method w.r.t. to the state-of-art coders.