Supersampling

Abstract: Surface elements (surfels) are a powerful paradigm to efficiently render complex geometric objects at interactive frame rates. Unlike classical surface discretizations, i.e., triangles or quadrilateral meshes, surfels are point primitives without explicit connectivity. Surfel attributes comprise depth, texture color, normal, and others. As a pre-process, an octree-based surfel representation of a geometric object is computed. During sampling, surfel positions and normals are optionally perturbed, and different levels of texture colors are prefiltered and stored per surfel. During rendering, a hierarchical forward warping algorithm projects surfels to a z-buffer. A novel method called visibility splatting determines visible surfels and holes in the z-buffer. Visible surfels are shaded using texture filtering, Phong illumination, and environment mapping using per-surfel normals. Several methods of image reconstruction, including supersampling, offer flexible speed-quality tradeoffs. Due to the simplicity of the operations, the surfel rendering pipeline is amenable for hardware implementation. Surfel objects offer complex shape, low rendering cost and high image quality, which makes them specifically suited for low-cost, real-time graphics, such as games.

Abstract: The rendering procedure used by neural radiance fields (NeRF) samples a scene with a single ray per pixel and may therefore produce renderings that are excessively blurred or aliased when training or testing images observe scene content at different resolutions. The straightforward solution of supersampling by rendering with multiple rays per pixel is impractical for NeRF, because rendering each ray requires querying a multilayer perceptron hundreds of times. Our solution, which we call "mip-NeRF" (a la "mipmap"), extends NeRF to represent the scene at a continuously-valued scale. By efficiently rendering anti-aliased conical frustums instead of rays, mip-NeRF reduces objectionable aliasing artifacts and significantly improves NeRF's ability to represent fine details, while also being 7% faster than NeRF and half the size. Compared to NeRF, mip-NeRF reduces average error rates by 17% on the dataset presented with NeRF and by 60% on a challenging multiscale variant of that dataset that we present. Mip-NeRF is also able to match the accuracy of a brute-force supersampled NeRF on our multiscale dataset while being 22x faster.

Abstract: In contrast to conventional multipixel cameras, single-pixel cameras capture images using a single detector that measures the correlations between the scene and a set of patterns. However, these systems typically exhibit low frame rates, because to fully sample a scene in this way requires at least the same number of correlation measurements as the number of pixels in the reconstructed image. To mitigate this, a range of compressive sensing techniques have been developed which use a priori knowledge to reconstruct images from an undersampled measurement set. Here, we take a different approach and adopt a strategy inspired by the foveated vision found in the animal kingdom—a framework that exploits the spatiotemporal redundancy of many dynamic scenes. In our system, a high-resolution foveal region tracks motion within the scene, yet unlike a simple zoom, every frame delivers new spatial information from across the entire field of view. This strategy rapidly records the detail of quickly changing features in the scene while simultaneously accumulating detail of more slowly evolving regions over several consecutive frames. This architecture provides video streams in which both the resolution and exposure time spatially vary and adapt dynamically in response to the evolution of the scene. The degree of local frame rate enhancement is scene-dependent, but here, we demonstrate a factor of 4, thereby helping to mitigate one of the main drawbacks of single-pixel imaging techniques. The methods described here complement existing compressive sensing approaches and may be applied to enhance computational imagers that rely on sequential correlation measurements.

Abstract: Nonuniform sampling of images is a useful technique in computer graphics, because a properly designed pattern of samples can make aliasing take the form of high-frequency random noise. In this paper, the technique of nonuniform sampling is extended from two dimensions to include the extra parameter dimensions of distribution ray tracing. A condition for optimality is suggested, and algorithms for approximating optimal sampling are developed. The technique is demonstrated at low sampling densities, so the characteristics of aliasing noise are clearly visible. At supersampling rates, this technique should move noise into frequencies above the passband of the pixel-reconstruction filter.