光譜算圖之進階採樣方法

Abstract

隨著科技的發展，無論在遊戲或是模擬或是室內設計上，算出一張擁有正確光影特效的圖成為一件非常重要的事情。傳統上使用如path tracing等等可以收斂出正確結果的演算法往往會需要大量的時間。隨著GPU性能的增加，人們認為在不久的將來就可以見到遊戲等應用使用相關的技巧做及時的計算。然而，傳統的演算法如path tracing並沒有考慮到光的頻率波長等等，因此傳統上使用這類的方法無法得到包含色散等等的結果。雖然將演算法調整以支援波長的採樣並不是一件難事，然而，若是加入了波長相關的計算會因為多一個維度採樣的原因使得整體結果更加難以收斂。本篇論文的主要目標便是透過各種採樣的技巧嘗試降低波長採樣帶來的收斂減速效果，使得我們可以在一個合理的時間內得到一張包含色散效果的相對沒有太多雜訊的圖。With the rapid development of technology, it has been an important thing to generate a photorealistic image with correct lighting effect whether in games, simulations or interior design. Traditionally, we often uses algorithms such as path tracing in order to generate a photorealistic image. However, it often takes a lot of time in order to generate a converged image without noise. As the performance of GPU increases year by year, it is believed that we can see real time unbiased photorealistic rendering used in applications like games in near future. However, Traditional algorithms such as path tracing didn't take into account the frequency or wavelength of light. As a result, images generated by these methods would not include lighting effects such as dispersion. Although it is not a big problem to adjust the algorithm to consider wavelength, it would take a lot more time to converge because of the additional dimension involved in the integration. The main goal of this paper is to increase the speed of convergence when we add wavelength sampling using various method. This makes us generate a converged image with dispersion effect without too much noise in a much more reasonable time.