摘要
为了保护数据隐私,私密图像在上传到云服务器之前需要进行加密,然而,加密会导致传统的图像特征无法提取,进而给图像检索带来困难。针对密文图像的检索问题,提出了一种云环境下JPEG图像的安全检索方法:数据拥有者部分解码JPEG码流得到图像的DCT(discrete cosine transform)系数,对系数进行置乱加密然后生成密文图像并上传到云服务器;云服务器在图像密文上提取DC系数差分特征以及LBP(local binary patterns)特征,通过比较图像的特征向量之间的距离来确定图像的相似度,最后返回相似图像。该方法不仅减少了计算复杂度,而且使得数据拥有者与云服务器之间的交互次数尽可能地减少,同时,与其他几种目前有代表性的方法相比,具有更好的安全性和检索准确度,能实现对JPEG图像安全高效的检索。最后基于该方法做了简单的仿真系统,进一步验证了该方法的有效性。
In order to protect data privacy, image with sensitiveor private information needs to be encrypted before being outsourced to the cloud. However, this leads to difficulties in image retrieval. This paper proposed a secure and efficient image retrieval method. Scrambling encryption encrypted images on the DCT domain, and outsourced the encrypted images to the serv- er. Then the method extracted the DC difference histogram and LBP among the image blocks as image feature vectors by the server. Whether the images were similar depend on the distance between feature vectors, based on the distances, the serverreturned the similar images to the user. Both of the image confidentiality and retrieval accuracy were ensured in the proposedmethod with less computational complexity and conmmunication cost compared with the other methods. The proposed method is particularly suitable for JPEG image retrival under cloud environment. Finally, this paper built a simulation system based on the proposed method and it works.
出处
《计算机应用研究》
CSCD
北大核心
2017年第4期1239-1243,共5页
Application Research of Computers
基金
国家"973"计划资助项目(2011CB302204-5)
国家自然科学基金资助项目(41571426)
