The Research On Image Vectorization Based On Diffusion Curve

With the rapid development of computer software and hardware, imageprocessing technology has been widely used in many areas of society. Dot rasterimage analysis, recognition and reconstruction technique in which the image vectorgraphic objects are also widely used in engineering, such as drawings, maps and otherscanning input, CNC cutting, carving and so on. However, due to the complexity ofthe image, any image vector algorithm has some limitations, lead to the use ofvectorization of image is limited in other fields. Therefore, the imagevectorization algorithm is still evolving, this technology still has a broad researchspace.This paper mainly discussed the curve color diffusion (diffusion curve)vectorization methods. First, make the color images into grayscale, then use edgedetection method to obtain information of the Image contour, save contourinformation (such as coordinates, color values, etc.) based on the original color image,Finally, according to information obtained by the contour original image, Jacobiiterative method is used to restore The original image. The main work is as follows:Firstly, we improved the method of extracting image contour. Canny,a classicalmethod,which was used to extracting image contour in original method.however,thismehod first make a color image into grayscale and then computer the Gradient valueof Image pixels,the gradient values obtained can not fully reflect the gradient valuesof the pixels in the color image, sometimes,a complete edge is not obtained. Weuse the method which was used to solve the tensor gradient to computer the gradientvalues of each pixel in the image,then select the edge point by the gradient value.Byusing the improved method,we can get a more complete and concise picture.Secondly, we improved the algorithm of restoring image.In order to obtain thecolor values on both sides of the image contour, the original method need Bezierfitting twice,then use the algorithm of rapid generation Voronoi graph to renderVoronoi diagram.and computer the nearest distance between the contour point andnon contour point,finally,use Jacobi iteration method to restore the image.The methodfitting the Bezier curves twice, and rendering the Voronoi diagram twice, we improvethe method by using the expansion method in Mathematical morphology.we only needto fit Bezier curve and rendering a Voronoi one time to restore the original image.Finally,based on the shortcomings that the original algorithm can not restore theimage which color was distributed Uneven (even if you can restore the original image, the effect is far from ideal), we put forward a method which give importance tocontour point to change the effects of certain other non-point profile contour points,thereby restoring the uneven color distribution picture.