My renderer update – Photon-Mapping(caustics) & Texture

本文链接：https://blog.youkuaiyun.com/saintony/article/details/5786881

本文介绍了作者在两周内为GI渲染器完成的两个关键组件：纹理和光线映射（Caustics）。使用DevIL进行纹理加载并实现Mipmap功能。通过光子映射实现Caustics效果，并采用网格加速技术提高渲染效率。

摘要生成于 C知道，由 DeepSeek-R1 满血版支持，前往体验 >

In recent two weeks, I finished another two important components of a GI renderer: caustics and texture.

[Texture]: I use DevIL as the basic image loading library. DevIL also offers Mipmap building function. I adopt it directly and what I have implemented is to choose the most proper size sub-image

[Photon-Mapping (Caustic)]: The main reference is the seminal paper (Jenson, 1996). It requires another individual pass of photon projection from lights to objects. The photons will follow the regular paths of regular rays but they will stop on the diffuse surfaces. Each photon contains information of color & position. In the rendering pass, extra color estimation will be evaluated according to the photon densities of each pixel using reconstruction kernel functions.

As (Jenson 1996) mentioned, caustic computation requires high photon number, because small objects may generate a large scale of caustic due to the curvature – especially when you need high quality caustics. But I use Grid Acceleration on caustics reconstruction, so the caustic rendering now keeps a const cost of each pixel computation because of the identical grid index reference actions (but of course photon grid organization costs more time.)

A sample screenshot (3x anti-aliased, about 123k photons):

Venus Caustics