卷积神经网络在数字图像隐写与识别中的应用

Application of Convolution Neural Network in Digital Image Recognition

基于卷积神经网络设计出一识别数字图像是否进行过隐写的算法。通过CNN中ResNet残差神经网络模型,主要针对于图像隐写算法中的LSB空域隐写算法进行图像隐写分类,ResNet神经网络在“残差块”之间的输出输入之间引入一个跳跃连接,缓解在网络加深过程的梯度消失问题,使得能通过结构的优化和深度的不断加深来提高其网络分类能力,最终基于ResNet建立对LSB隐写图像具有较高准确率的分类模型。

This paper designs an algorithm based on convolutional neural network to recognize whether the digital image has been steganographed.Through the ResNet residual neural network model in CNN,the LSB spatial steganography algorithm in the image steganography algorithm is mainly used for image steganography classification.The ResNet neural network introduces a jump connection between the output and input of the“residual blocks”to alleviate the ladder disappearance problem in the network deepening process,so that the network can be improved through the optimization of the structure and the deepening of the depth.Finally,a classification model based on ResNet with high accuracy for LSB steganographic images is established.