Image Texture Enhancement Method based on Convolutional Neural Network 

Pingping Zeng  

College of Science and Technology, Nanchang University, Nanchang, 330000, China 

Abstract: The existing image texture enhancement methods have the problem of imperfect image structure and texture layer decomposition process, resulting in low definition. A novel image texture enhancement me-thod based on convolutional neural network is designed. The local gray changes of the image are measured to obtain the texture mapping index of the image. The convolutional neural network decomparts the image struc-ture and texture layer to remove the information that does not conform to the given scale. By introducing the histogram matching constraint, the deblurring model is constructed to increase the gradient value of the image detail area and improve the image texture enhancement mode. Experimental results: The average sharpness of the image texture enhancement method in this paper and the other two image texture enhancement methods are 63.952, 53.340 and 54.952 respectively, indicating that the image texture enhancement method integrated with convolutional neural network has higher application value. 

Keywords: Convolutional neural network; Image texture; Feature fusion; The texture layer; Image edge; Brightness