Stitches Generation for Random-Needle Embroidery Based on Markov Chain Model.

We present a new stylization method to generate Random-needle Embroidery stitches which is a graceful Chinese embroidery art formed by intersecting stitches. First, we model the intersecting stitch and initialize stitches positions and directions for regions. The Markov chain model is used to select similar intersecting stitches for filling each region to avoid artifacts in local area. Then a hierarchical iterative stitches generation process is used to keep the characteristic of multi-layering of stitches. Finally, top layer stitches in each iteration of the generation process are slightly moved according to bottom stitches by a Lloyd's method based approach to make stitches maintain the characteristic of uniform distribution. Experiments show that our result stitches can avoid artifacts in local area and maintain the uniformity and multi-layering at the same time. Comparing with the state-of-the-art methods, our result stitches have a richer visual effect.