Google Inc.

Nvidia Corporation

Yun-Ta Tsai

Jonathan Tilton Barron

Tiancheng Sun

Stephen William Keckler

Paul E Debevec

Sofien Bouaziz

Dawid Stanislaw Pajak

Sean Ryan Francesco Fanello

Rohit Kumar Pandey

Christoph Rhemann

Markus Steinberger

Iuri Frosio

Sean Ryan Fanello

Chloe LeGendre

Tianfan Xue

Daniel Robert Johnson

Navin Sarma

Chih-Chi Cheng

Jason Lavar Clemons

Xiuming Zhang

Michael Milne

Apparatus and methods related to applying lighting models to images of objects are provided. An example method includes applying a geometry model to an input image to determine a surface orientation map indicative of a distribution of lighting on an object based on a surface geometry. The method further includes applying an environmental light estimation model to the input image to determine a direction of synthetic lighting to be applied to the input image. The method also includes applying, based on the surface orientation map and the direction of synthetic lighting, a light energy model to determine a quotient image indicative of an amount of light energy to be applied to each pixel of the input image. The method additionally includes enhancing, based on the quotient image, a portion of the input image. One or more neural networks can be trained to perform one or more of the aforementioned aspects.

Enhanced photo relighting based on machine learning models

Apparatus and methods related to applying lighting models to images of objects are provided. A neural network can be trained to apply a lighting model to an input image. The training of the neural network can utilize confidence learning that is based on light predictions and prediction confidence values associated with lighting of the input image. A computing device can receive an input image of an object and data about a particular lighting model to be applied to the input image. The computing device can determine an output image of the object by using the trained neural network to apply the particular lighting model to the input image of the object.

Photo relighting using deep neural networks and confidence learning

Examples relate to implementations of a neural light transport. A computing system may obtain data indicative of a plurality of UV texture maps and a geometry of an object. Each UV texture map depicts the object from a perspective of a plurality of perspectives. The computing system may train a neural network to learn a light transport function using the data. The light transport function may be a continuous function that specifies how light interacts with the object when the object is viewed from the plurality of perspectives. The computing system may generate an output UV texture map that depicts the object from a synthesized perspective based on an application of the light transport function by the trained neural network.

Neural light transport

Methods for white-balancing images are provided. These methods include determining, for an input image, a chrominance histogram for the pixels of the input image. The determined histogram is a toroidal chrominance histogram, with an underlying, toroidal chrominance space that corresponds to a wrapped version, of a standard flat chrominance space. The toroidal chrominance histogram is- then convolved with a fitter to generate a two-dimensional heat map that is then used to determine art estimated chrominance of i|lummaiioB present id the input image; This can include fitting a bivariate von Mises distribution, or some other circular and/or toroidal, probability distribution, to the determined two-dimensional heat map. These methods for estimating illumination chrominance values for input images have reduced computational costs and increased speed relative to other methods for determining image illuminant chrominance values.

Fast fourier color constancy

A computing device may obtain an input image. The input image may have a white point represented by chrominance values that define white color in the input image. Possibly based on colors of the input image, the computing device may generate a two-dimensional chrominance histogram of the input image. The computing device may convolve the two-dimensional chrominance histogram with a filter to create a two-dimensional heat map. Entries in the two-dimensional heat map may represent respective estimates of how close respective tints corresponding to the respective entries are to the white point of the input image. The computing device may select an entry in the two-dimensional heat map that represents a particular value that is within a threshold of a maximum value in the heat map, and based on the selected entry, tint the input image to form an output image.

Convolutional color correction in digital images

Hardware-based convolutional color correction in digital images

A patch memory system for accessing patches from a memory is disclosed. A patch is an abstraction that refers to a contiguous, array of data that is a subset of an N-dimensional array of data. The patch memory system includes a tile cache, and is configured to fetch data associated with a patch by determining one or more tiles associated with an N-dimensional array of data corresponding to the patch, and loading data for the one or more tiles from the memory into the tile cache. The N-dimensional array of data may be a two-dimensional (2D) digital image comprising a plurality of pixels. A patch of the 2D digital image may refer to a 2D subset of the image.

Patch memory system

A computer implemented method of performing an approximate-nearest-neighbor search is disclosed. The method comprises dividing an image into a plurality of tiles. Further, for each of the plurality of tiles, perform the following in parallel on a processor: (a) dividing image patches into a plurality of clusters, wherein each cluster comprises similar images patches, and wherein the dividing continues recursively until a size of a cluster is below a threshold value; (b) performing a nearest-neighbor query within each of the plurality of clusters; and (c) performing collaborative filtering in parallel for each image patch, wherein the collaborative filtering aggregates and processes nearest neighbor image patches from a same cluster containing a respective image patch to form an output image.

Efficient approximate-nearest-neighbor (ANN) search for high-quality collaborative filtering

Growing community of inventors

Santa Clara, CA, United States of America

Yun-Ta Tsai