Research On Raster Image Vectorization With Application

In recent years, the research subject on raster image to vector graphics conversion(R2V) is increasingly popular in the field of CAD & CG It applies to scanning and recognizing of engineering drawings, maps as well as NC cutting, carving. Embroideries have been favored by many people because of their special aesthetic and decorative features. embroidering with computer have emerged consequently. At this time, designment efficiency of sketch in the embroidery CAD system has become the bottleneck of embroideries manufacture efficiency. Presently, the embroidery CAD system can generate needle code automatically only if the vector graphics is provided. But the sketch is designed by the means of hand-drawn directly or outlining referenced image. Therefore it showed up the inefficiency of this way.In this dissertation, The scheme that embroidery sketch is designed rapidly is proposed,which is using raster image to vector graphics conversion. Several problems standing in need of solutions are investigated, which includes pretreatment of the black-white hand-drawn line image,segmentation of the color image that patterns by some big area color regions, points chain of target extraction and fitting. The main contributions of this dissertation are summarized as follows:Firstly, considering speckle noise,separation,adherence probably exist in the black-white hand-drawn line image ,the corresponding rapid and effective solutions are proposed.Secondly,the effective method for exatraction of point chain is provided through synthetizing related algorithm proposed by others. Preconcerted goal is carred out.Thirdly,Two method of segmentation for color image are presented according to the feature of color image. If the color of noise and the main color are similar,considering pattern apperceived by human vision associates not only with hue, but also with area and location of color regions,the two-step method,that is firstly color clustering,secondly regions combination is proposed.For other cases,the interactive segmentation method based on region growing is introduced.Experiment results show that both efficency and correctness are gived attention,and the image segmented is available to designment of embroidery sketch.Fourthly, the method of fitting lines and arcs to data points directly based on genetic algorithm(GA) is adopted .The parameters and steps of algorithm adopted are gived. The test results prove that this method is not sensitive to noise measured and the location of key points is exact.The prominent idea is that the scheme of color image segmentation firstly color clustering and then regions combination is proposed.The description and computing method of relationship between two regions in image space are provided. They are successfully applied in automatic combination of some small area noise regions and interactive combination of some big area regions.Thus the segmentation of color image is rapid and efficient.