The improved algorithm for color-texture image segmentation

This paper present an improved algorithm based seeded region growing (SRG) for color-texture textile image segmentation. The algorithm improve the over-segmented phenomenon of the color-texture textile image used Euclidean distance, relative Euclidean distance and discrete cosine transform (DCT) to quantify the color and texture information in seed selection, region growing and region merging steps. Experimental results show that the algorithm has good segmented results for color-texture textile image in the case of selecting appropriate parameters.