Design of Oil Painting Damaged Area Repair Algorithm Based on Convolution Neural Network

Abstract

Oil painting has been introduced into China for nearly a century. In recent decades, the introduction of the concept and technique of oil painting in Europe has lagged far behind, but the introduction of the concept and technique of modern painting has lagged behind. Moreover, oil painting has rich strokes and forms of expression, and is good at expressing charm in an abstract way. This kind of painting art works is quite different from the conventional natural scene images in vision. The traditional oil painting damaged area repair algorithm is too sensitive to black and white color when calculating the pixels in the damaged area, which is prone to color coverage. Therefore, we propose an algorithm design of oil painting damaged area restoration based on convolutional neural network. The convolution neural network is used to repair the damaged oil painting. The restoration process is modeled by the restoration function, and the restoration model is constructed by applying additive noise to the input oil painting image with damaged areas. The core idea of the algorithm is to extract the candidate areas of seals or inscriptions in Chinese painting and calligraphy images by using the prior knowledge of the scene (such as the color and edge of seals and inscriptions). Then, the pre-trained model is used to accurately locate the seal or inscription.