Improvement of foreground extraction of garment image based on skin color detection combined with Grab Cut

In order to solve the problem of the interference of the models skin color and the under-segmentation of some complex images when the automatic Grab Cut algorithm is applied to foreground extraction of garment images, a new algorithm of foreground extraction is proposed. In order to remove the skin area of the model, an improved skin detection algorithm based on bilateral filtering for image denoising is proposed. In order to remove the under-segmentation regions and preserve the foreground regions of a garment, a contour detection algorithm is proposed. The improved skin color detection method and the contour detection algorithm are combined with the automatic Grab Cut algorithm to extract the foreground of garment images. The experimental results show that, compared with the other existing methods, the improved skin color detection method increases the accuracy of skin detection significantly and that the proposed foreground extraction algorithm is better than the automatic Grab Cut algorithm in garment foreground extraction.