Region growing segmentation for textile yarn images

This paper addresses the problem of segmenting textile yarn images. The aim of image segmentation in the considered application is to extract fibers protruding from the yarn core. Region-growing based approach is proposed for yarn image segmentation. The method is guided and constrained by coherence enhancing diffusion (CED) filter. Results of image segmentation provided by the proposed method are presented and compared with results provided by the traditional clustering approaches. The comparison proves that the proposed segmentation algorithm provides high quality results. The number and the length of fibers extracted from the background are sufficient for determination of yarn properties performed in the following steps of the measurements.