Feature fusion-based two-stage generative adversarial network digital mural restoration method

• Main Authors: ZHAO NA, SHI RUOLIN, YANG YU, LI HUILI, SHEN QI, LYU QIONGSHUAI, GAO JINGLI, GONG YUEHONG
• Format: Patent
• Language: Chinese; English
• Published: 14.05.2024
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; IMAGE DATA PROCESSING OR GENERATION, IN GENERAL; PHYSICS

The invention relates to a feature fusion-based two-stage generative adversarial network digital mural restoration method. The method comprises the following four parts: constructing a model training data set, constructing a feature fusion two-stage generative adversarial network model, training the feature fusion two-stage generative adversarial network model and restoring a damaged mural. The method comprises a coarse-grained generation network, a fine-grained generation network and a discrimination network. And feature extraction and feature fusion are performed by using a feature fusion module, and the reconstruction performance of the model is improved by using a two-stage training strategy. The method comprises the following specific steps: inputting a damaged mural image and a line draft image into a trained feature-fused two-stage generative adversarial network model, performing complementation and color filling on a damaged area by a coarse-grained generative network, and generating a preliminarily r