Encoding of still pictures by a wavelet transform and singular value decomposition

This paper proposes a new scheme for picture compression which decomposes a picture to (3/spl times/3+1) subband pictures using the 2-D discrete wavelet transform three times, and then decomposes a picture block matrix A(16/spl times/16 pixels) of the lowest frequency subband picture to singular values and eigenvectors of AA/sup T/ and A/sup T/A and encode them by PCM. For all other subband pictures, first we encode the 0-level wavelet coefficient run-length by a modified Huffman code and then encode the non-0-level coefficients by PCM with non-linear quantizer.