Segmentation of Color Textures Using K-Means Cluster Based Wavelet Image Fusion
This paper develops a new method for cluster segmentation of color texture image. At first, a color texture image has been decomposed three sub-bands that R (red), G (green), B (blue) channel. Second, each sub-bands channel has been decomposed by discrete wavelet transform (DWT) to the same resolution. Third, fusion operation is performed on the transformed images according to PCA- based (Principal Component Analysis) weighted average rule to choose the reconstructed coefficients. Then all the decomposed sub-bands images are reconstructed using inverse discrete wavelet transform (IDWT) that becoming a fusion image. Finally, k-means cluster segmentation algorithm has been applied for this fusion image, the segmentation results has been obtained. A number of experiments are performed to demonstrate the efficacy of the proposed method.
Z. K. Huang et al., "Segmentation of Color Textures Using K-Means Cluster Based Wavelet Image Fusion", Applied Mechanics and Materials, Vols. 20-23, pp. 209-214, 2010