Abstract：The paper presents an algorithm based on classification and 3D-SPIHT for 1–7 bands multispectral images compression. First, the 3D transform for the image data is used with CDF97 wavelet to move spacial correlation and classification prediction to move spectral correlation. Second, according to remain image coefficients from classification prediction, the method realizes compression coding by 3D-SPIHT and adopts Huffman coding files to do the lossless coding of codebook and index table. Meanwhile it transmits all coding files to the decoding so as to reconstruct image. The results of numerous experiments indicate the quality of reconstructive image.
2006, Vol. 28 
