Matting algorithm and application based on Mahalanobis distance and the fuzzy C-means clustering algorithm

Based on Mahalanobis distance and the fuzzy C-means algorithm, this article introduces a digital color image matting algorithm. First the red, green and blue color components of color image pixels are normalized. Second the appropriate mask as a sample set is selected in the background of the normalized image, and the Mahalanobis distance between each pixel and the sample set is calculated. Third the calculated Mahalanobis distances are classified into two categories using the fuzzy C-means clustering algorithm:the foreground and the background. Finally, the quality of the matting is improved using the filling-hole technique. Eight images have been processed for comparison, the results show that this algorithm can automatically segment these images, and is better than the Mahalanobis distance algorithm, fuzzy C-means clustering algorithm and linear regression algorithm.