Nm. Nasrabadi et Sa. Rizvi, NEXT-STATE FUNCTIONS FOR FINITE-STATE VECTOR QUANTIZATION, IEEE transactions on image processing, 4(12), 1995, pp. 1592-1601
In this paper, the finite-state vector quantization scheme called dyna
mic finite-state vector quantization (DFSVQ) is investigated with rega
rd to its subcodebook construction. In DFSVQ, each input block is enco
ded by a small codebook called subcodebook which is created from a muc
h larger codebook called supercodebook. Each subcodebook is constructe
d by selecting, using a reordering procedure, a set of appropriate cod
e-vectors from the supercodebook. The performance of the DFSVQ depends
on this reordering procedure; therefore, several reordering procedure
s are introduced and their performances are evaluated in this paper. T
he reordering procedures that are investigated, are based on the condi
tional histogram of the code-vectors, index prediction, vector predict
ion, nearest neighbor design, and frequency usage of the code-vectors.
The performances of the reordering procedures are evaluated by compar
ing their hit ratios (the number of blocks encoded by the subcodebook)
and their computational complexity. Experimental results are presente
d and it is found that the reordering procedure based on the vector pr
ediction performs the best when compared with the other reordering pro
cedures.