Mural Restoration Research Based on Samples and Deep Learning

Mural paintings are an important part of cultural heritage, but due to the wind and rain of time, most of the murals have been damaged and faded to various degrees. How to restore and protect murals has become a hot research topic. The traditional mural restoration method has some defects, such as difficulty in controlling the accuracy and quality of restoration, great damage to the original cultural relics, and other problems, and the evaluation is unreasonable. Therefore, this paper proposes a sample and deep learning algorithm to evaluate and analyze the quality of mural restoration based on the creation. Firstly, the damaged images and sample images are preprocessed according to the repair quality requirements, including denoising and alignment, to reduce the interference factors in the repair quality evaluation. Then, the deep learning algorithm is used to extract the features of damaged images and sample images to form a mural restoration quality evaluation scheme, and the mural restoration quality evaluation results are comprehensively analyzed. The experimental results show that under the condition that the evaluation criteria are fixed, the restoration effect of the deep learning algorithm on the damaged mural is better than the traditional mural restoration method in terms of repair accuracy and restoration time.