Chinese traditional landscape painting super-division method based on diffusion model

• Main Authors: ZHAO NA, YANG YU, SHEN QI, LYU QIONGSHUAI, GAO JINGLI, GONG YUEHONG
• Format: Patent
• Language: Chinese; English
• Published: 15.12.2023
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; IMAGE DATA PROCESSING OR GENERATION, IN GENERAL; PHYSICS

A Chinese traditional landscape painting super-division method based on a diffusion model comprises the following steps: constructing a model training data set, constructing a Chinese traditional landscape painting super-division diffusion model, constructing a Chinese traditional landscape painting super-division back diffusion denoising reconstruction process, and performing color correction on the obtained super-resolution Chinese traditional landscape painting by using a color correction method. And the final high-resolution Chinese traditional landscape painting is obtained. According to the method, a diffusion model is used for improving the resolution reconstruction problem in the traditional Chinese landscape painting digitization process, noise is gradually added to an input image by constructing a forward Gaussian diffusion process, the training process of a noise estimation network is constrained by using a loss function, and after the model training is completed, the resolution reconstruction prob