Chinese Painting Style Transfer Based On Convolutional Neural Network

Image style transfer refers to learning the style of a painting and then applying this style to another picture.Style transfer extracts the style semantics of famous paintings,and expresses the style features of the image with data,so as to modify the style of another picture,modify the image features,and realize the style transfer.Image style transfer plays an important role in image restoration and image classification,and is an important research direction in computer image processing.In recent years,with the development of neural networks,many image style transfer methods based on neural networks have been proposed.However,when these methods are applied to Chinese painting,the effect is not satisfactory.The reason is that traditional Chinese painting has a unique style description.Thousands of years of sinking make Chinese traditional painting unique.In order to effectively solve the problem of style transfer of traditional Chinese painting,this article conducts research on the basis of existing methods to explore style transfer methods that are more suitable for traditional Chinese painting.The main work includes the following:1.Taking the performance of Chinese traditional painting in the deep learning characteristics as a starting point,I have studied the differences between Chinese and Western paintings.By observing and comparing Chinese paintings with Western paintings and natural pictures,analyzing the characteristics of Chinese paintings from the perspective of painting techniques The line,texture and structure of the image are quite different from other images.2.Propose a method of Chinese painting image style transfer based on feature reorganization.Using the relationship between the deep features in the neural network and the image semantics,a method of expressing the artistic style of traditional Chinese painting is proposed,which is combined with existing methods to match the feature matrix in the production network through feature reorganization,and finally use decoding The generator generates the migration image.The experimental results prove that the method based on feature reorganization can improve the results of style transfer of Chinese painting to a certain extent.3.Propose a style transfer method of Chinese painting image based on generative confrontation network.The cyclic generative confrontation network is improved,the texture of the target image is input as an additional condition into the generator,and the discriminator is changed to a relativistic discriminator to build a new generative confrontation network.In order to prove the effectiveness of the method,a large number of Chinese ink landscape paintings and natural landscape photos are used as experimental data.The experimental results show that the generated images obtained after the style transfer have improved authenticity and quality.