An improved image texture synthesis based on algorithm convolution neural network

Abstract

In recent years, texture synthesis technology has made great progress and has become one of the research hotspots in the fields of computer graphics, computer vision and image processing. Texture synthesis is proposed to solve problems such as seam aliasing in texture mapping. With the development of artificial intelligence technology, texture synthesis methods can achieve more effective results. Therefore, this paper improves the traditional texture synthesis method based on convolutional neural network. By using the VGGnet network model of the convolutional neural network and adding the Batch Normalization (BN) layer to individual convolutional layer affiliated to the network, the network training speed is improved. In this paper, the source texture image and an image initialized by white noise are respectively input into the convolutional neural network. After the forward propagation process, the texture features of the images are extracted, the loss function is constructed, and then the network is trained to update the composite image. Extensive results of experiments indicate that our proposed approach outperforms other texture synthesis methods in terms of image synthesis quality.