基于多种检测像素预测方式的LSB图像信息隐藏检测

本文突破了传统LSB隐写分析算法只针对图片整体特征提取、监测手段不够精细的缺陷,提出了针对每一像素点的针对性判断。算法主要利用图像函数估计、噪声判断、像素值估计等方法,通过预测每一像素是否被改动,来判断整张图片的嵌入情况,从而实现小嵌入率隐写分析与嵌入点大致估计的目标。算法的检测率较以往算法有明显提高,对低嵌入率（1%以内）的LSB嵌入检测效果尤为有效。同时对高嵌入率的嵌入,算法保证了99.9%以上的准确率。

Detection of LSB Image Steganalysis Based on Multi Pixel Predicting Methods

Aim at the drawback of the traditional LSB（Least Significant Bit） steganalysis algorithm,which could only extract the feature of a whole picture with a unsatisfying monitoring accuracy,we put forward a corresponding judgement about each image pixel.Our algorithm mainly used the ways of image function estimation,noise detection and pixel estimation to predict whether one pixel is modifi ed,afterwards we can determine the embedment of the whole picture.After the analysis,we trained threshold to achieve the goal of an approximate estimate of a embedding point with a low embedding rate steganalysis pursuantly.In terms of experiment results,our algorithm has better performance than the traditional algorithm,especially in the low embedding rate situation（the rate is less than 1%）.For the situation of high embedding rate（the rate is less than 10%）,the accuracy of our algorithm can be more than 99.9%.