Image Vectorization Based On TCB-splines

The vector representation of image is scalable,compact and resolution indepen-dent and it facilitates many image editing processes.Therefore,various vector graphics primitives have been proposed to represent images,which include lines,curves and polygons.However,existing graphics-based image representation-s still face technical challenges,these methods are not equipped to handle the complexity of color or curvilinear features with arbitrary shapes and orientations in many images.A powerful vector representation called TCB-splines that is able to effectively approximate raster images and their curvilinear features as well.We create dis-continuous basis pairs across the feature on the basis of collinear knots or multiple knots,the control points corresponding to these basis have different color value while sharing same position,which leading to the image crease across the feature.An automatic algorithm is developed for approximation framework which convert a raster image to vector image based on the TCB-splines of degree two represen-tation.Firstly,we improve feature extraction method to acquire accurate feature points and then parameterize these points into poly-lines constructed adaptively by the curvature of feature points.Secondly,appropriate distribution of knots is generated through variational mesh generation approach with feature preserva-tion.Then we construct discontinuous basis on feature edge through constrained LTP procedure and fit target image iteratively.Compared with existing vector-based image representations,the proposed algorithm has the extra advantages in representing feature region with continuity of C-1 while other area keep the con-tinuity of C1.The effectiveness of the proposed method has been demonstrated in experiments.