Image upsampling using one or more neural networks

Apparatuses, systems, and techniques are presented to reconstruct one or more images. In at least one embodiment, one or more neural networks are used to upsample one or more images based, at least in part, on one or more brightness values. 402-) 1 Receive data for a current image in a sequence 404-...

Full description

Saved in:
Bibliographic Details
Main Authors Cohen, Loudon, Janis, Pekka, Massal, Gregory
Format Patent
LanguageEnglish
Published 18.05.2023
Subjects
CALCULATING
COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS
COMPUTING
COUNTING
IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
PHYSICS
Online AccessGet full text

Cover

More Information
Summary:Apparatuses, systems, and techniques are presented to reconstruct one or more images. In at least one embodiment, one or more neural networks are used to upsample one or more images based, at least in part, on one or more brightness values. 402-) 1 Receive data for a current image in a sequence 404-LI Calculate an auto exposure value for this current image Use auto exposure value in tonemapping to reduce a dynamic range of this current image and a prior image in sequence Process current and prior image, with reduced dynamic range, using a neural network to infer pixel-specific blending weights 1 L Generate, using blending weights, a blended output image at lower dynamic range Use calculated auto exposure value to reverse tonemap blended output image back to higher dynamic range 4144 414-L, Provide output image at least to application, for presentation, or for blending a subsequent image
Bibliography:Application Number: AU20220259808