Correlation Alignment Total Variation Style Transfer Model Based On Tensorflow And Grab Cut

The style transfer of digital images is one of the research hotspots in the field of deep learning in recent years and one of the very interesting applications in the field of machine vision.The style transfer model based on the convolutional neural network has attracted extensive attention of researchers due to the good artistic effect of the resulting image.The image style transfer technology can transfer the style of the style image to the content image,so that the resulting image contains both the main content structure information of the content image and the style information of the style image,to meet people's requirements.However,the resulting image of the model has problems such as uneven style texture,noise enhancement,and long model iteration time,which affects the final effect of style transfer.And the model can only transfer the texture information of one style image to one content image at a time,which greatly reduces the enthusiasm of the user to create the style transfer result image.Aiming at the above-mentioned problems,this paper has carried out in-depth research on it based on convolutional neural networks.The main research contents and innovations in the article are as follows.This paper introduces the related theories of image style transfer.First,it introduces the various stages experienced in the development of neural networks.Secondly,the mathematical derivation process of the artificial neural network,the convolutional neural network and the VGG network used in the model mentioned in the paper,the total variation regular denoising model and its solution process are introduced.Next,the image style transfer model based on convolutional neural network is explained.Taking the traditional style transfer model as an example,the article expounds the method of extracting the content information and style texture used in the classic style transfer model,and the specific process of the model's style transfer.Finally,the TensorFlow deep learning framework used to implement the model mentioned in the article is introduced.First,a new model of total variational style transfer based on correlation alignment is proposed.Secondly,by analyzing and comparing the reconstruction results of different convolutional layers after CNN decomposes the image,a new convolutional layer selection strategy is proposed.Then,the parameters of the proposed new model are determined through experiments.Finally,the proposed model is compared with the traditional image style transfer method through multiple methods such as result image,experiment time,total loss value and subjecti-ve evaluation.Experimental results show that the proposed model is superior to the classic style transfer model in the visual effect of the resulting image and the running efficiency of the algorithm.The article first explains the problems of traditional image style transfer algorithms.In order to effectively solve these problems,the Grab Cut algorithm for image segmentation and the color harmony algorithm for optimizing the image tone are introduced,and then the implementation method and related loss function of the style transfer model based on Grab Cut are introduced.Finally,a large number of comparative experiments illustrate the effectiveness of the method,and it can be preliminarily concluded that the method has a certain universality.