Browsing by Author "Chang, Chun-Fa"

Now showing 1 - 2 of 2
  • Thumbnail Image
    Item
    Specular Manifold Bisection Sampling for Caustics Rendering
    (The Eurographics Association and John Wiley & Sons Ltd., 2022) Jhang, Jia-Wun; Chang, Chun-Fa; Umetani, Nobuyuki; Wojtan, Chris; Vouga, Etienne
    We propose Specular Manifold Bisection Sampling (SMBS), an improved version of Specular Manifold Sampling (SMS) [ZGJ20]. SMBS is inspired by the small and large mutations in Metropolis Light Transport (MLT) [VG97]. While the Jacobian Matrix of the original SMS method performs well in local convergence (the small mutation), it might fail to find a valid manifold path when the ray deviates too much from the light or bounces from a complex surface. Our proposed SMBS method adds a large mutation step to avoid such a problematic convergence to the local minimum. The results show SMBS can find valid manifold paths in fewer iterations and also find more valid manifold paths. In scenes with complex reflective or refractive surfaces, our method achieves nearly twice or more improvement when measured in manifold walk success rate (SR) and root mean square error (RMSE).
  • Thumbnail Image
    Item
    An Unbiased Hybrid Rendering Approach to Path Guiding
    (The Eurographics Association, 2021) Jhang, Jia-Wun; Chang, Chun-Fa; Bittner, Jirí and Waldner, Manuela
    When we think of hybrid rendering of rasterization and ray tracing, we often consider rasterization as a mean to solve the primary rays and then consider ray tracing as a mean to add secondary effects. We take a different approach to combine ray tracing and rasterization, in which our final output images are still produced with a path tracer. We leverage the GPU rasterization to build the necessary data that are required for path guidng, thus improves the convergence of our path tracer. We borrow the ideas of Voxel Cone Tracing and implement it in GPU shaders to build the path guiding data. The advantage of our proposed hybrid approach is that it maintains the unbiased results of a Monte Carlo path tracer while incurring relatively small performance hit in path guiding, as shown in our preliminary results.