Pixar

Pixar Animation Studios

Anthony David DeRose

Michael Kass

Mark Joseph Meyer

Steve Bako

Tien Gia Truong

The present disclosure relates to using a neural network to efficiently denoise images that were generated by a ray tracer. The neural network can be trained using noisy images generated with noisy samples and corresponding denoised or high-sampled images (e.g., many random samples). An input feature to the neural network can include color from pixels of an image. Other input features to the neural network, which would not be known in normal image processing, can include shading normal, depth, albedo, and other characteristics available from a computer-generated scene. After the neural network is trained, a noisy image that the neural network has not seen before can have noise removed without needing manual intervention.

De-noising images using machine learning

A method for determining interior coordinates is disclosed. The method includes receiving information specifying an object having a plurality of sites and a boundary. Interior coordinates associated with each of the plurality of sites are determined based on the boundary. The interior coordinate associated with each of the plurality of sites represent a system of coordinates that satisfy several properties, including non-negativity and interior locality. At least one value associated with the plurality of sites is then interpolated using the interior coordinates.

Techniques for interior coordinates

Different limit surfaces are derived from the same initial arbitrary polygon mesh by sequentially combining different subdivision rules. This added freedom allows for the more efficiently modeling of objects in computer graphics including objects and characters with semi-sharp features.

Hybrid subdivision in computer graphics

Realistic computer generated character animation is achieved by modeling skin or other covering as a separate a quasi-static surface elastically tied to an underlying detailed, articulated model controlled by the animator. The result, automatically incorporates realistic stretching, and sagging, and the combined effect of multiple overlapping animation controls while retaining for the animator frame-by-frame control of a detailed model of the character.

Realistic surface simulation in computer animation

Methods for defining smooth scalar fields on arbitrary polygonal meshes which can be consistently carried through the subdivision process involve computing mesh vertices after subdivision to a suitable resolution and computing scalar field values at the vertices corresponding to an arbitrary smooth surface, including the limit surface. The methods allow one to define and compute the value of scalar fields over an arbitrarily dense set of points on the limit surface. Applications of these techniques in computer graphics and computer animation include: (1) the definition of pseudo-coordinates for use in parametric shading including texture mapping; (2) the assignment of smoothly varying articulation weights over the control points of a model to more efficiently define the transformation of a highly detailed object or character under specific animation controls; and (3) the assignment of smoothly varying local parameters for specifying the local behavior of dynamic or quasi-static objects or surfaces.

Growing community of inventors

San Rafael, CA, United States of America

Anthony David DeRose