电气设备红外图像可逆信息隐藏方法研究

发布时间：2018-03-14 20:55

本文选题：电气设备　切入点：红外图像　出处：《华北电力大学(北京)》2017年博士论文　论文类型：学位论文

【摘要】：基于热红外图像的故障分析和诊断技术是电力设备状态检修的重要手段。基于数字图像的可逆信息隐藏方法可以将图像产生的现场信息以不可见的方式嵌入其中,在需要的时候进行信息提取,以实现对海量电气设备热红外图像的本源性认证。本文以现场采集的工况电气设备图像为研究对象,对可逆信息隐藏关键技术进行了深入研究,主要研究工作如下:为了解决工况电气设备红外图像椒盐噪声多,分布随机性大,造成压缩后的奇异像素位置图过长,占据大量嵌入空间的问题,提出了高密度跳变位置图调整方法。该方法通过修改短连续零组,主动舍弃部分跳变频繁的可扩展预测误差,只保留跳变强度低于某阈值的可扩展预测误差,提高了位置图中相同属性标记的连续性,压缩了大量位置图冗余,提高了水印净荷负载的嵌入容量。针对逐像素预测算法中单零预测误差扩展的容量限制,借鉴固定分块方法中的排序预测和误差扩展技术,提出了融合型预测误差扩展方法。新算法设计正负两个扩展方向,在每一个扩展方向上都确定一个扩展标准,使得两个方向的扩展策略独立运行,提高了嵌入容量。针对当上下文像素较多时,局部粗糙度无法有效描述目标像素的预测区域内部的像素分布细节特征,提出了基于极值深度和局部粗糙度的联合像素选择算法。该方法利用预测误差多重直方图,优先选择高平滑像素进行扩展,实现预测误差扩展的精细化选择。实验结果表明,所提算法提高了峰值信噪比。针对单层信息嵌入时,嵌入信息比特对像素的修改方向是单一的,提出了基于部分抵消的双预测器双层可逆信息隐藏方法。新方法赋予了第二层嵌入时对第一层嵌入时的部分像素进行反方向修改的能力,这些像素在历经两次信息嵌入后实现了效果抵消,被还原为信息隐藏前的状态,使得同一个像素不仅最多可以携带2比特信息,提高嵌入容量,还可以降低由此产生的附加失真。
[Abstract]:Fault analysis and diagnosis technology based on thermal infrared image is an important means of condition maintenance of power equipment. Reversible information hiding method based on digital image can embed the scene information of image in invisible way. In order to realize the original authentication of mass thermal infrared images of electrical equipment, the information extraction is carried out when needed. In this paper, the key technology of reversible information hiding is deeply studied, taking the electrical equipment images collected on the spot as the research object. The main research work is as follows: in order to solve the problem that the infrared image of electrical equipment has a lot of pepper and salt noise, large distribution randomness, resulting in the compression of the singular pixel location map is too long, occupying a large amount of embedded space. In this paper, a method for adjusting the position diagram of high density jump is proposed. By modifying the short continuous zero group, the extensible prediction error with frequent partial hopping is abandoned actively, and only the extensible prediction error with a jump intensity lower than a certain threshold is retained. The continuity of the same attribute mark in the location map is improved, the redundancy of the location map is compressed, and the embedding capacity of the watermark payload is improved. Based on the technique of ranking prediction and error spreading in the fixed block method, a new algorithm is proposed, in which a new algorithm is designed for both positive and negative expansion directions, and an extension criterion is determined for each extension direction. The extended strategy in both directions can run independently and improve the embedding capacity. When the context pixels are more the local roughness can not effectively describe the pixel distribution details inside the prediction region of the target pixel. A joint pixel selection algorithm based on extreme depth and local roughness is proposed, in which high smooth pixels are preferentially selected for expansion using multiple histograms of prediction errors. The experimental results show that, The proposed algorithm improves the peak signal-to-noise ratio (PSNR). A double-layer reversible information hiding method based on partial cancellation is proposed, which gives the ability to modify the reverse direction of some pixels in the first layer when the second layer is embedded. After twice embedding, these pixels are reduced to the state before information hiding, so that the same pixel can not only carry up to 2 bits of information, but also increase the embedding capacity. Additional distortion resulting from this can also be reduced.
【学位授予单位】：华北电力大学(北京)
【学位级别】：博士
【学位授予年份】：2017
【分类号】：TP391.41;TP309

【相似文献】