02320nas a2200361   4500000000100000000000100001008004100002653002300043653004700066653002300113653001600136653003500152653003400187653002400221653002400245653001200269653001100281653002900292653001800321653002600339653002100365653003600386653001800422653001200440100001200452700001100464245010200475856015300577300001200730490000700742520118900749022002001938                                       d10aImage segmentation10aImage segmentation and feature extractions10aImage stylizations10aImage-based10aImage-based artistic rendering10aIntangible cultural heritages10aIntersection angles10aLinear combinations10aNeedles10aPixels10aRandom-needle embroidery10aRasterization10aSparse representation10aTextile printing10aVector and pixel space depended10aVector spaces10aVectors1 aK. Yang1 aZ. Sun00aVector and Pixel Space Depended Stitch Definition and Style Transfer for Random-Needle Embroidery  uhttps://www.scopus.com/inward/record.uri?eid=2-s2.0-85053155010&doi=10.3724%2fSP.J.1089.2018.16471&partnerID=40&md5=af95a8664d680df1e7a1c470851b891b  a778-7900 v303 aWe present an image stylization method to simulate Random-needle Embroidery which has been designated as Intangible Cultural Heritage. We first define the rendering primitive of this art, namely, stitch, i.e., collection of threads arranged in a certain pattern inseparably both in vector space and pixel space in vector space, we design different thread distributions which include the orientations, intersection angles, lengths and widths of threads in stitches, and in pixel space, the rendering primitives are generated by rasterizing the stitches on 2D pixel arrays. We also generate a dictionary to represent the generated stitches. Afterwards, for an input image, based on image segmentation and feature extraction, each region is partitioned into sub-regions to describe the content of the region. Finally, for each sub-region new stitches can be synthesized by optimizing a linear combination of stitch dictionary atoms via sparse representation, and rendering image is generated by placing stitches with different attributes on the canvas. Experimental results show that our method can effectively translate the input image into an art image with the style of random-needle.  a10039775 (ISSN)