Font Size: a A A

Image Vectorization Based On Feature Edges Extraction

Posted on:2016-08-07Degree:MasterType:Thesis
Country:ChinaCandidate:R WuFull Text:PDF
GTID:2308330470467672Subject:Computer technology
Abstract/Summary:PDF Full Text Request
Image vectorization is an important research field in digital image processing, and can be widely used in cartoon animation, art creation, online image display etc. With the emerges of various display devices and screen resolutions, the vectorized representation of images has drawn researcher’s intense interests due to its characteristics of low storage consumption, resolution independence and geometric editability. The kernel of image vectorization is to find a continuous geometry and color representation for a discrete image. It is a challenge topic. Among various methods, the diffusion curves approach has advantages in ease of editing and keeping color gradient, but may lead to leak of color if detected edges are not closed. Besides, the method includes intensive interactions, thus it’s very time-consuming.An automatic image vectorization method is proposed in this paper, where closed feature edges can be extracted from segmented image. Thus it solves the color leaking problem in the aforementioned methods. First, input image is preprocessed to reduce its noises and enhance its characteristics, so as to facilitate regional clustering segmentation. Features of different color regions after clustering segmentation are then marked, and closed feature edges of the image are extracted based on a 2D multi-material marching squares method. To guarantee geometric topology of the image as well as simplify further management and operations, the discrete feature edges data is fitted constrainedly with piecewise cubic Bezier curves. Finally, all the curves and color primitives are stored as a vector file. Color diffusion gradient is computed through solving a Poisson equation with constraints. Color gradient effect of vectorized image is obtained by vector rendering using multi-grid iterations. As a result, the automatic vectorization is achieved without color leaks and tedious manual interactions.
Keywords/Search Tags:Image vectorization, Color feature, Boundary extraction, Curve fitting, Possion equation
PDF Full Text Request
Related items