基於擴散曲線之點陣圖自動向量化

Abstract

擴散曲線是一種新的向量基元，這種基元被繪製時會將圖片上的區域分開，而且可以在曲線兩側定義顏色。這些顏色會從曲線上平滑地向兩側擴散，遂至整張圖片。本論文基於擴散曲線的特性，提出了一個點陣圖向量化的自動化方法。我們從點陣圖中萃取輪廓、顏色及模糊資訊，並轉換成擴散曲線所需要的幾何基元。我們的方法可以將點陣圖轉換成一組擴散曲線，這組擴散曲線所表示的向量圖會很接近原始的點陣圖內容，而且可以很容易地編輯或製作成動畫。

Diffusion curve is a novel vector-based primitive. It partitions the space through which it is drawn and define colors on each side. These colors smoothly diffuse outwards from each side until they cover the entire image. In this thesis, we propose a method to automatically vectorize a raster image. Our method extracts the contour, color and blur attribute from the raster image, and represents the raster image by the geometry primitives of diffusion curves. The vector graphics, which is represented by a set of diffusion curves closely approximates the original image, can be easily edited or animated.