| Hi-index | 0.00 |
In this paper, an unsupervised parallel approachcalled Fuzzy Competitive Learning Network (FCLN) forVector Quantization (VQ) and Spread FCLN (SFCLN) forcolor image compression in the Mean value / Differencevalue Transform (MDT) domain are proposed. In theFCLN, the codebook design is conceptually considered asa clustering problem. Here, it is a kind of competitivelearning network model imposed by the fuzzy clusteringstrategies working toward minimizing an objectivefunction defined as the average distortion measurebetween any two training vectors within the same class.The color image information transformed by MDToperation was separated into RGB 3-plane mean valueand detail coefficients. Then the detail coefficients for eachplane were trained using the proposed SFCLN method togenerate the VQ codebook. The experimental results showthat promising codebooks can be obtained using theproposed FCLN and SFCLN for color image compressionin the MDT domain.