Image Enhancement Using Normalizing Flows

Number: 20220005161
Published: 2022-01-06
Filed: 2020-07-01
Assignee: Disney Enterprises, Inc.
Inventors: Djelouah; Abdelaziz, Helminger; Leonhard Markus, Bernasconi; Michael, Schroers; Christopher Richard
CPC: G06N3/045; G06N3/0455; G06N3/0475; G06N3/08; G06T3/4053; G06T5/60; G06T5/70; G06T5/77
Verdict: Set aside generic image enhancement
Source: Google Patents · FreePatentsOnline

Abstract

According to one implementation, an image enhancement system includes a computing platform including a hardware processor and a system memory storing a software code configured to provide a normalizing flow based generative model trained using an objective function. The hardware processor executes the software code to receive an input image, transform the input image to a latent space representation of the input image using the normalizing flow based generative model, and perform an optimization of the latent space representation of the input image to identify an enhanced latent space representation of the input image. The software code then uses the normalizing flow based generative model to reverse transform the enhanced latent space representation of the input image to an enhanced image corresponding to the input image.

Background

BACKGROUND

Image restoration and image enhancement have seen significant progress due to recent developments in the field of deep neural networks. Nevertheless, most conventional techniques rely on the availability of training data in the form of pairs of images with and without degradation. As a result, the applicability of conventional image restoration and image enhancement techniques is limited to use cases in which training data can be obtained and in which the type of image degradation to be reversed can be identified in advance. Unfortunately, for historically valuable or artistically significant legacy images, that information is often impossible to obtain due to unavailability of non-degraded original imagery. Thus, there is a need in the art for image enhancement solutions that do not require foreknowledge of the type or the extent of degradation that an image has undergone.

Claims

1. An image enhancement system comprising: a computing platform including a hardware processor and a system memory; a software code stored in the system memory, the software code configured to provide a normalizing flow based generative model trained using an objective function; the hardware processor configured to execute the software code to: receive an input image: transform, using the normalizing flow based generative model, the input image to a latent space representation of the input image; perform an optimization of the latent space representation of the input image to identify an enhanced latent space representation of the input image; and reverse transform, using the normalizing flow based generative model, the identified enhanced latent space representation of the input image to an enhanced image corresponding to the input image. || 11. A method for use by an image enhancement system including a computing platform having a hardware processor and a system memory storing a software code configured to provide a normalizing flow based generative model trained using an objective function, the method comprising: receiving, by the software code executed by the hardware processor, an input image; transforming, by the software code executed by the hardware processor and using the normalizing flow based generative model, the input image to a latent space representation of the input image; performing, by the software code executed by the hardware processor, an optimization of the latent space representation of the input image to identify an enhanced latent space representation of the input image; and reverse transforming, by the software code executed by the hardware processor and using the normalizing flow based generative model, the identified enhanced latent space representation of the input image to an enhanced image corresponding to the input image.