基于深度学习的动画图像纹理细节增强方法研究  

Research on the Texture Detail Enhancement Method of Animation Images Based on Deep Learning

作　　者：蔡希阳 CAI Xiyang(Quanzhou Normal University,Quanzhou Fujian 362000,China;Graduate University of Mongolia,Ulaanbaatar,Mongolia)

机构地区：[1]泉州师范学院,福建泉州362000 [2]蒙古国研究大学,蒙古乌兰巴托016199

出　　处：《佳木斯大学学报(自然科学版)》2024年第12期33-36,共4页Journal of Jiamusi University:Natural Science Edition

摘　　要：为提高图像的清晰度与用户体验,引进深度学习算法,设计动画图像纹理细节增强方法。使用滑动窗口计算熵值,将熵值归一化作为权重;利用卷积神经网络(CNN)的多层非线性变换逐步抽象特征信息;使用图像分割技术,将图像分为前景和背景两部分,根据计算得到的图像背景散射程度,进行图像纹理细节的增强处理。实验结果表明:设计方法可以确保构建的动画图像与场景的实际帧数达到设计帧数,即确保场景的流畅度与动画清晰度。In order to improve the clarity and user experience of the image,the deep learning algorithm is introduced to design the texture detail enhancement method of the animated image.The sliding window is used to calculate the entropy,and the entropy is normalized into the weight.The feature information is abstracted step by step by using multi-layer-nonlinear transformation of convolutional neural network(CNN).Using image segmentation technology,the image is divided into foreground and background.According to the calculated background scattering degree,the image texture details are enhanced.The experimental results show that the design method can ensure that the actual frame number of the constructed animation image and scene can reach the design frame number,that is,to ensure the fluency of the scene and the clarity of the animation.

关 键 词：深度学习 权重项 增强方法 动画图像 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]