Well, it differs from typical SSAO in two ways: first, the sampling points are chosen completely in 2D, no 3D-2D reprojection. second, each sample is treated as a small disk surface, we extract its normal form the normal buffer and we reconstruct a view space position from the depth buffer and screen coordinates. Using these disks and the disk to disk occlusion formula contained in the nvidia paper, a much more realistic image results.
Well, it differs from typical SSAO in two ways: first, the sampling points are chosen completely in 2D, no 3D-2D reprojection. second, each sample is treated as a small disk surface, we extract its normal form the normal buffer and we reconstruct a view space position from the depth buffer and screen coordinates. Using these disks and the disk to disk occlusion formula contained in the nvidia paper, a much more realistic image results.
(chana, es pa quien lo entienda me temo xD)
arkano22 2 years ago