Title:
TRAINING METHOD FOR IMAGE PROCESSING NETWORK, AND CODING METHOD, DECODING METHOD, AND ELECTRONIC DEVICE
Document Type and Number:
WIPO Patent Application WO/2024/032075
Kind Code:
A1
Abstract:
Provided in the embodiments of the present application are a training method for an image processing network, and a coding method, a decoding method and an electronic device. The training method comprises: acquiring a first training image and a first predicted image, and acquiring a period of a checkerboard artifact, wherein the first predicted image is generated by means of performing image processing on the first training image on the basis of an image processing network; according to the period, dividing the first training image into M first image blocks, and dividing the first predicted image into M second image blocks, wherein the sizes of the first image blocks and the sizes of the second image blocks are all related to the period; and then, determining a first loss on the basis of the M first image blocks and the M second image blocks; and then, training the image processing network on the basis of the first loss. After an image processing network is trained on the basis of the training method in the present application, a checkerboard artifact in an image, which is obtained by processing performed by the trained image processing network, can be eliminated to a certain extent, thereby improving the visual quality of the image, which is obtained by processing performed by the trained image processing network.
Inventors:
LI MENG (CN)
WANG JING (CN)
SHI YIBO (CN)
WANG JING (CN)
SHI YIBO (CN)
Application Number:
PCT/CN2023/095130
Publication Date:
February 15, 2024
Filing Date:
May 18, 2023
Export Citation:
Assignee:
HUAWEI TECH CO LTD (CN)
International Classes:
G06N3/08; G06N3/04; G06T9/00
Foreign References:
| CN107369189A | 2017-11-21 | |||
| CN110473141A | 2019-11-19 | |||
| CN113469897A | 2021-10-01 | |||
| US20220189116A1 | 2022-06-16 |
Attorney, Agent or Firm:
BEIJING RUN ZEHENG INTELLECTUAL PROPERTY LAW FIRM (CN)
Download PDF: