Title:
IMAGE RECONSTRUCTION MODEL TRAINING METHOD AND APPARATUS, DEVICE, MEDIUM, AND PROGRAM PRODUCT
Document Type and Number:
WIPO Patent Application WO/2023/216720
Kind Code:
A1
Abstract:
The present application relates to the technical field of image processing and discloses an image reconstruction model training method and apparatus, a device, a medium, and a program product. The method comprises: acquiring a first sample image and at least two types of second sample images (302); respectively adding damage features corresponding to the at least two types of second sample images into the first sample image to generate at least two types of single degraded images (304); fusing the at least two types of single degraded images to obtain a multiple degraded image corresponding to the first sample image (306); performing image reconstruction processing on the multiple degraded image to generate a predicted reconstructed image corresponding to the multiple degraded image (308); calculating a loss function value on the basis of the second sample images, the single degraded images, the first sample image, and the predicted reconstructed image (310); and updating a model parameter of a model on the basis of the loss function value (312). According to the model trained by means of said method, accumulative errors caused by sequential image reconstruction are avoided.
More Like This:
Inventors:
HUANG YAWEN (CN)
ZHENG YEFENG (CN)
ZHANG LE (CN)
ZHENG YEFENG (CN)
ZHANG LE (CN)
Application Number:
PCT/CN2023/082436
Publication Date:
November 16, 2023
Filing Date:
March 20, 2023
Export Citation:
Assignee:
TENCENT TECH SHENZHEN CO LTD (CN)
International Classes:
G06N3/08; G06V10/774; G06V10/80
Foreign References:
| CN115115900A | 2022-09-27 | |||
| CN109191411A | 2019-01-11 | |||
| CN109146813A | 2019-01-04 | |||
| CN111311704A | 2020-06-19 | |||
| CN112529776A | 2021-03-19 | |||
| US20140375636A1 | 2014-12-25 |
Attorney, Agent or Firm:
BEIJING SAN GAO YONG XIN INTELLECTUAL PROPERTY AGENCY CO., LTD. (CN)
Download PDF: