基于局部相关性的LSB匹配隐写分析

针对LSB匹配嵌入机制，提出一种利用局部区域像素之间相关性的LSB匹配隐写分析方法。根据较高位平面灰度之间的距离选择图像的平坦区域，将随机性较强的最低两个和三个位平面取出，统计相同灰度值图像块的个数，使用中值滤波对图像进行滤波复原处理，得到差异特征。引入和图像的均值进一步提高检测性能，共计得到4维特征使用Fisher线性分类器判别载体、载密图像。实验证明，该方法对未压缩高精度图像具有较好的检测性能。

Steganalysis for LSB matching based on correlation of pixels in local regions

Spatial LSB matching steganography decreases the correlation between pixels. Based on this, a method was presented based on the changes of correlation in local regions. The smooth areas were selected according to the range of the gray values in the higher bit planes, then the numbers of the image blocks were counted, which were with the same gray level in the least two and three bit planes. Make the image filtered by Median Filter to get the feature for detection. The average of the sum image was introduced for further improvements and 4 features were extracted in all to detect. Fisher linear discriminator was applied for classifying. Experimental results show that the proposed method exhibits excellent performances for the detection of LSB matching steganography in grayscale images.