基於柏林雜訊之可控制隨機性材質生成

Abstract

柏林雜訊廣泛運用在產生自然現象，例如：雲，火焰，木紋和地形等應用。藉由加入柏林雜訊增加擾動也可增添動畫的變化度。然而，過去較少的研究是針對控制柏林雜訊。我們提出一個可以修改和控制柏林函數所生成的雜訊值，並且不會破壞原有雜訊函數所擁有的統計上的特性。我們可以產生非常貼近使用者需求的雜訊樣式並且保有原本雜訊函數該有的統計特性。我們將處理的問題運用多層式最佳化方法，而分層的依據及最佳化順序是由低頻率雜訊到高頻率雜訊。我們的方法可以輕易的達到整體性或是區域性的控制例如程序化材質樣式控制，並且我們可以有效率的再產生相似雜訊樣式。

Perlin noise is widely used to render natural phenomena such as cloud, fire, woodgrain and terrain. It is also used to enrich the variety of animation by adding disturbance to the motion of objects. However, there is less attention on controlling Perlin noise in the past. We present a new approach to modify and control the noise value generated by Perlin noise function without destroying the statistical properties of the noise function. We can generate the controllable noise that closely matches a user’s demand pattern while preserving the original statistical properties of noise. We formulate our problem as a multi-level optimization problem in which the optimization process is executed from low frequency level to high frequency level. Our approach can easily achieve global and local control on texture pattern design and can reproduce the same pattern efficiently. Keyword: