A MASSIVELY PARALLEL ALGORITHM FORVECTOR QUANTIZATION

1 Introduction This paper is concerned with the parallel implementation of full-search vector quantizer systems on a MasPar MP-2, a single-instruction, multiple-data (SIMD) massively parallel computer. Vector quantization [1] is a very effective, but computationally complex method for data compression. The simplest form of a vector quantizer operates as follows. Blocks of k consecutive samples of the input waveform are grouped together to form k-dimensional vectors. The input vectors are compared with each of the vectors in the codebook and a metric or distance is computed in each case. The representative vector closest to the input vector is identified and its index is made available to the output stage. In this paper, we use the Euclidean distortion measure given b