摘要
图像的二次编辑和篡改大大降低了图像的可信度,受夜间光线影响,图像分类过程不稳定,图像来源取证准确率较差,为此,研究基于迁移卷积神经网络的夜间低照度图像同源性鉴定算法。基于迁移卷积神经网络,将夜间低照度图像分类,计算不同特征图像之间的相关性;增强与降维处理夜间低照度图像,为图像同源性鉴定提供标识性依据;通过划分图像块,实现夜间低照度图像同源性鉴定。经实验论证分析,应用算法能使夜间低照度图像内容更为完整,细节显示更多,图像同源性鉴定准确率和类别纯度更高,具有有效性与实用性。
The secondary editing and tampering of the image greatly reduce the credibility of the image.Affected by the night light,the image classification process is unstable,and the accuracy of image source forensics is poor.Therefore,the homology identification algorithm of nighttime low illumination images based on the migration convolution neural network is studied.Based on the migration convolution neural network,the nighttime low illumination images are classified,and the neuron correlation between different characteristic images is calculated.It enhances and reduces the dimension of the nighttime low illumination images,and provides the identification basis for the identification of image homology.By dividing the image blocks,the homology identification of nighttime low illumination images is realized.After the experimental demonstration and analysis,this algorithm is applied,the content of nighttime low illumination images is more complete,the details are displayed more,the accuracy of image homology identification and the purity of categories are higher,and it is effective and practical.
作者
庞展
张旭毅
PANG Zhan;ZHANG Xuyi(Judicial Police Academy,Xinjiang University of Political Science and Law,Tumushuke 843806,China;Henan Police College,Zhengzhou 462000,China)
出处
《微型电脑应用》
2024年第4期178-181,共4页
Microcomputer Applications
关键词
迁移卷积神经网络
夜间低照度图像
图像增强
同源性鉴定
噪声特征
图像分块
transferred convolutional neural network
nighttime low illumination image
image enhancement
homology identification
noise characteristic
image segmentation
