An iteratively interpolative vector quantization algorithm for image data compression

An iterative scheme is proposed in which vector quantization is applied to images represented by interpolative pyramids. An N×N image, where N=2ⁿ+1, is subsampled successively to form a pyramid. A `spline under tension' interpolation algorithm is applied to each level to form an interpolative pyramid. A differential pyramid is then generated by taking the difference between the corresponding levels of the subsampled pyramid and the interpolative pyramid. A universal code book can be found to vector quantize the differential pyramid on a level-by-level basis and the pyramid can be transmitted iteratively. The receiver recovers the information by decoding the differential pyramid and adding the difference into the interpolative pyramid at each level. The algorithm properly carries out the errors of vector quantization to the next level of the pyramid