三維中國水墨畫之快速合成演算法

Abstract

水墨畫在中國傳統文化藝術中是具有相當悠久歷史的一項藝術，使用墨與水調和出各種不同濃淡的墨色，用墨色表現出的明暗對比來描繪出一幅景物。在現今已經發展出來的許多模擬中國水墨畫的技術中，我們可以看到許多不錯的結果。但是因為演算法以及硬體的限制，執行速度不夠快，所以在某些應用程式上鮮少被使用到。本篇論文研究的重點，便是改善水墨畫效果產生的運算速度。在本篇論文中，我們使用硬體語言來撰寫，利用GPU來運算，將不同的影像技術加入模擬程式中，最後將系統產生的結果利用渲染的技術，建立一個自動產生水墨畫效果的即時互動式系統。

The ink painting is an art with long history in Chinese culture. By mixing water and ink properly, we can obtain different concentrations to show the contrast of an object. Several techniques focused on simulating Chinese ink paintings can yield good results. But they can not have good efficiency when used in some applications because of the limitation in hardware and algorithm. In this thesis, the major contribution is to improve the performance of synthesizing the Chinese ink styles. We use each random particle which is generated in 3D scene to draw billboard and use High-Level Shader Language (HLSL) to write our program code which is running in the GPU. We build a system which puts different image space techniques in shader program and mix them to generate a scene in Chinese ink painting style automatically.