Research Of Image Vectorization

Image vectorization is one of the important problems in digital image processing, and it is a across subject including comprehensive computer vision, computer image processing, computer graphics and intelligence, and other disciplines. Vector image has a lot of advantages:first of all, vector image is made of simple geometric figure components, with compact structure and small storage space; Secondly, the vector image is easy to edit, when we edit vector image, such as rotating, stretching, we only need to modify the corresponding parameter information of geometric components; third, the vector object is easy to enlarge or compression, and its display quality on the computer will not be reduced, magnification of the vector image can keep sharp features of the horn side, and won’t appear fuzzy and influence display quality.For a given color raster image, we can get pixels information of the image, pixel information including position information and color information (usually using RGB value).Change the input color images to vector image, we can easily to realize the extension and compression, editing, animation of the image, and other operations. Many existing image vectorization algorithms only apply to the engineering drawings, and in recent years the image vectorization algorithms began to deal with photos of level color images.This paper first studied image vectorization using optimization gradient meshes. For a color image, we segment the image into sections firstly, processing the image with some child objects, and then using quadrilateral grid to approximate objects. Gradient grids structured by this method are Ferguson surfaces in essence. Due to join the gradient information, the method has very good effect to tonal color image.This article proposed two kinds of new image vectorization method based on partial differential equation’ analytical solution and numerical solution, extracting information of pixel line by line, calculating the control points of b-spline through information of these pixels, and the b spline curves are elliptic partial differential equation’s boundary conditions. Image vectorization method based on partial differential equation’s analytical solution efficiently calculate partial differential equation’s solution through analytical solution method, and reconstruct the whole image through the generated patches. Image vectorization method based on partial differential equation’s numerical solution calculate partial differential equation’s solution through finite difference method, and get the pixel information of adjacent point, and then get information of other unknown pixels through the bilinear interpolation method to reconstruct the whole image.Two kinds of new image vectorization method based on partial differential equation’ analytical solution and numerical solution process image using the method of processing graphics, and the principle of vectorization is simple, very efficient and convenient for a simple image.