A Competitive Learning Algorithm with Controlling Maximum Distortion
Takeshi Miura, Kentaro Sano, Kenichi Suzuki, and Tadao Nakamura
Graduate School of Information Sciences, Tohoku University, 6-6-01 Aramaki Aza Aoba, Aoba-ku, Sendai 980-8579, Japan
Vector quantization with an optimal codebook is attractive for lossy data compression. So far, a number of codebook design algorithms have been proposed to minimize the mean square error, MSE. However, these algorithms have a problem that MSE minimization sometimes causes an unacceptable maximum-distortion, which is very important in several applications. This paper proposes a competitive learning algorithm with controlling maximum distortion that designs a codebook giving a maximum distortion within a given error bound. The proposed algorithm assigns a code vector to an input vector with a too large distortion. Experimental results showed that the algorithm has both maximum-distortion control and MSE minimization capability.
This article is published under a Creative Commons Attribution-NoDerivatives 4.0 Internationa License.