摘要： 基于像素值排序的可逆数据隐藏算法通过修改图像块中的最大像素和最小像素嵌入数据，但并未充分利用图像块内的每一个像素，从而影响嵌入性能。结合块再分原理，提出基于像素值排序的可逆数据隐藏算法。将原始图像划分为3×3的图像块，计算每一个图像块的局部复杂度。设计12种分块模式将局部复杂度小于阈值的图像块细分为子块A和B。根据子块A和B的不同局部特征分别采用2种不同的扫描顺序读取像素。子块A的像素序列使用次小值预测最小值和次大值预测最大值的方法，获得2个预测误差值，子块B的像素序列利用中值像素连续预测其余4个像素的方法，得到4个预测误差值。在此基础上，利用图像块中预测误差值为0和1的像素嵌入隐藏数据。实验结果表明，该算法在一个图像块中最高可嵌入6 bit的数据，在较低计算复杂度的情况下能够有效提高像素的嵌入性能。

关键词: 像素值排序, 可逆数据隐藏, 嵌入性能, 局部复杂度, 阈值

Abstract: The reversible data hiding algorithm based on Pixel Value Ordering (PVO) embeds data by modifying the largest and smallest pixels in the block without completely using each pixel in image block, thereby affecting the embedding performance.Combined with the principle of block subdivision, this study proposes an reversible data hiding algorithm based on PVO by dividing the original image into 3×3 image blocks to calculate the local complexity of each.Twelve block modes are constructed for subdividing the image block, whose local complexity is less than the threshold, into sub-blocks A and B.According to the different local features of sub-blocks A and B, two different scanning sequences are used to read pixels.In the pixel sequence of sub-block A, two prediction error values are obtained using the second smallest pixel to predict the minimum pixel and the second largest pixel to predict the maximum pixel.In the pixel sequence of sub-block B, four prediction error values are obtained by continuously predicting the remaining four pixels using the median pixel.On this basis, hidden data are embedded using pixels with prediction error values of zero and one in the image block.Experiments reveal that the algorithm can embed up to six bit of data in an image block and effectively improve the embedding performance of pixels with a low computational complexity.

Key words: Pixel Value Ordering (PVO), reversible data hiding, embedding performance, local complexity, threshold

中图分类号: 

• TN918