An FPCA-Based Color Morphological Filter for Noise Removal

Morphological filtering is a useful technique for processing and analysis of binary and grayscale images. The extension of morphological techniques to color images is not a straightforward task, because this extension stems in the multivariate ordering problem. Since the multivariate ordering is ambiguous, the existing approaches have used the known vector ordering schemes for the color ordering purpose. In the last decade, many different color morphological operators have been introduced in the literature. Some of them have focused on noise suppression purposes. However, none has shown good performance specially on edgy regions. In this paper, new color morphological operators based on fuzzy principle component analysis are proposed for the noise removal purpose. These operators employ statistical information (obtained by applying a fuzzy clustering algorithm on the color space) to achieve desired results for the denoising application. The performance of our proposed operators is compared with the recent morphological operators reported in the literature for denoising purpose and the superiority of the proposed method is shown.