Mip-NeRF: A Multiscale Representation for Anti-Aliasing Neural Radiance Fields
This work addresses aliasing issues in NeRF rendering for computer vision and graphics applications, offering a more efficient and accurate method, though it is incremental as it builds directly on NeRF.
The paper tackles the problem of aliasing and blurring in neural radiance fields (NeRF) due to single-ray rendering, by introducing mip-NeRF, a multiscale representation that renders anti-aliased conical frustums, which reduces error rates by 17% on standard datasets and 60% on a multiscale variant while being 7% faster and half the size.
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.