Abstract: This study proposes a novel method to segment color images based on new clustering algorithm and a new feature space obtained by transformation of RGB space using Fuzzy Principal Component Analysis (FPCA) technique. The clustering algorithm is based on recursive one-dimensional histograms analysis, that separate multimodal 1D-histograms. We propose an intuitive idea for selecting the fuzzy principal components, simultaneously with clustering task, not necessary the first principal components. Quantitative and qualitative analysis of the performance of the proposed approach is examined on both real and medical images. Some experiments show that segmentation is better when using FPCA technique.
H. Essaqote , N. Zahid , I. Haddaoui and A. Ettouhami , 2007. Color Image Segmentation Based on New Clustering Algorithm and Fuzzy Eigenspace . Research Journal of Applied Sciences, 2: 853-858.