Graphics Reference
In-Depth Information
Chen et al. also took advantage of epipolar geometry [Chen et al. 11]. They
noticed that they could accelerate the ray marching process by using a 1D min/-
max binary tree constructed for each epipolar slice. Their algorithm is rather
sophisticated. It relies upon a singular value decomposition of the scattering
term, and requires special care of the area near the epipole.
In this chapter we apply our previous approach [Yusov 13] to rendering scat-
tering effects in the Earth's atmosphere. We reduce the number of samples for
which the airlight integral is numerically integrated with the epipolar sampling
and exploit 1D min/max binary trees to accelerate ray marching. We also discuss
practical details, like integration with cascaded shadow maps [Engel 07].
2.3 Algorithm Overview
The following is a high-level overview of our algorithm. The remaining sections
provide details on each step. The algorithm can be summarized as follows:
1. Generate epipolar sampling.
(a) Compute entry and exit points of each epipolar line on the screen.
(b) Distribute samples.
2. Select ray marching samples.
(a) Compute coarse in-scattering for each sample.
(b) Sparsely locate initial ray marching samples along epipolar lines.
(c) Place additional samples where coarse in-scattering varies notably.
3. Construct 1D min/max binary tree for each epipolar slice.
4. Perform ray marching.
5. Interpolate in-scattering radiance for the rest of the samples from ray march-
ing samples.
6. Transform scattering from epipolar coordinates to screen space and combine
with the attenuated back buffer.
2.4 Light Transport Theory
Since our method ultimately boils down to eciently solving the airlight integral,
it requires an introduction. We will start with the key concepts of the light
transport theory.
