METHOD, APPARATUS AND SYSTEM FOR IMAGE-TO-IMAGE TRANSLATION

The method utilizes a GAN with coherent cycle learning and Gaussian fusion to translate RGB images into UV images, addressing the challenge of effective UV image synthesis for medical diagnostics, achieving high-resolution and realistic results with improved inference speed.

FR3146532B1Active Publication Date: 2025-03-28LOREAL SA
0 Cites 0 Cited by

Patent Information

Application Number
FR2023002159
Authority / Receiving Office
FR · FR
Patent Type
Patents
Current Assignee / Owner
Filing Date
2023-03-09
Publication Date
2025-03-28
Estimated Expiration
2043-03-09

AI Technical Summary

Technical Problem

Current methods struggle to effectively translate RGB images into ultraviolet (UV) images, particularly in medical diagnostics where UV images are useful for skin analysis and lesion detection.

Method used

A method and system using a generative adversarial network (GAN) with a coherent cycle learning approach to translate RGB images into UV images, employing Gaussian fusion to merge overlapping plots and reduce pixel discrepancies.

Benefits of technology

The approach achieves high-resolution, realistic UV image synthesis with high correspondence to real UV images, reducing grid effects and color inconsistencies, while also accelerating inference time.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

METHOD, APPARATUS, AND SYSTEM FOR IMAGE-TO-IMAGE TRANSLATION. A method, apparatus, and system, according to embodiments, provide image-to-image translations such as the synthesis of ultraviolet (UV) images from input images in an RGB (red, blue, green) color model. In one embodiment, a machine-learned generator produces overlapping UV patch images from overlapping RGB patch images extracted from an input image. The overlapping UV patch images are blended using a Gaussian weighting factor applied to overlapping pixels having the same location in the input image. The Gaussian blending distributes the pixel weights according to the pixel's distance from the center of its patch, with the weighting being highest at the center. Figure for abstract: none
Need to check novelty before this filing date? Find Prior Art