Non-Parametric Vector Quantization Algorithm
| Paper File | Download Paper File | | Appear In | ECTI Transaction CIT (ECTI Transaction CIT) | | Publication Date | 01/05/2007 - 30/05/2007 | | Volume | 3 | | Pages | 31 - 38 | | No | 1 | | Author 1 | Haemwaan Sivaraks | | Author 2 | Athasit Surarerks |
Abstract
Recently researches in vector quantization domain
concentrate on improving the performance of parametric
algorithms (i.e., need to specify either the
codebook size or the expected distortion.) The specification
of the both parameters gives the users some
difficulties. We, in the research, propose a nonparametric
vector quantization (NVQ) algorithm.
The concept is to manage Voronoi regions with respect
to the distortion decreasing rate. Experimental
results point that the average distortions are statistically
decreased comparing with LBG and adaptive
incremental LBG algorithms at the 99% of confidence
level. Both speech and image data are also included
in our consideration. |
