Vectorization Of Abstractive Images

Image abstraction is an important technique in digital image processing. It is widely applied in cartoon animation, computer art. two-dimensional game, advertising creation production and Internet applications. How to transfer, process and display the abstractive images on/among devices with different resolutions is an important problem. Image vectorization is an alternative solution to this problem since it is based on the geometrical description and has the advantages of compact description, storage and free of image resolution. In this thesis, the input image is first abstracted based on shape simplification and extracted feature flows. Then the abstracted image is quantized. The feature edge of quantized image is extracted and the color gradient is recorded for vectorization representation. Meanwhile, an appropriate rendering algorithm is designed to rasterize the vectorized image. The main contributions of the thesis are:1. It introduces a vectorization algorithm for abstractive images. The vectorized image has abstraction style and is independent of image resolution. Thus and the algorithm, reduces the information redundancy. There is no tedious manual interaction in the algorithm.2. An edge-based image vectorization model is proposed. To represent different regions in the abstraction image effectively, a closed boundaries extraction algorithm is designed and there is no gap between adjacent regions. With the help of the boundary information, the color information in the regions and Poisson diffusion equation, the image vectorization is achieved.