An Affinity Propagation Based method for Vector Quantization Codebook Design

TL;DR

This paper introduces an improved affinity propagation algorithm integrated with LBG for vector quantization codebook design, resulting in higher quality codebooks with better convergence than traditional methods.

Contribution

The paper proposes a modified affinity propagation algorithm combined with LBG, enhancing convergence and codebook quality in vector quantization.

Findings

Enhanced convergence ability of the proposed method

Generation of higher-quality codebooks

Outperforms conventional LBG in experiments

Abstract

In this paper, we firstly modify a parameter in affinity propagation (AP) to improve its convergence ability, and then, we apply it to vector quantization (VQ) codebook design problem. In order to improve the quality of the resulted codebook, we combine the improved AP (IAP) with the conventional LBG algorithm to generate an effective algorithm call IAP-LBG. According to the experimental results, the proposed method not only enhances the convergence abilities but also is capable of providing higher-quality codebooks than conventional LBG method.