基于离散小波变换和感知哈希的加密医学图像检索算法

针对加密存储在云服务器的医学图像安全检索问题，提出基于离散小波变换（DWT）和感知哈希的加密医学图像检索算法。首先结合Henon映射的特点对图像进行频域加密运算；然后，对加密医学图像进行小波分解，得到逼近原图的子图；其次，根据离散余弦变换（DCT）的特性，通过比较DCT各系数与系数均值的关系得到图像的感知哈希序列；最后通过比较感知哈希序列之间的归一化相关系数来实现对加密医学图像检索。与基于非负矩阵分解（NMF）的哈希算法相比，所提算法在高斯噪声下检索精度提高了近40%，且在JPEG压缩攻击、中值滤波攻击、缩放攻击和扭曲攻击下检索精度与之相差无几。实验结果表明，所提算法对于常规攻击和几何攻击具有较好的鲁棒性，同时降低了图像加密的时间复杂度。

Encrypted image retrieval algorithm based on discrete wavelet transform and perceptual hash

Focusing on medical image secure retrieval in cloud server, an encrypted medical image retrieval algorithm based on Discrete Wavelet Transform (DWT) and perceptual hash was proposed. Firstly, the image was encrypted in frequency domain based on the characteristics of Henon mapping. Secondly, the encrypted medical image was decomposed by wavelet to obtain the sub-image close to the original image. According to the characteristics of Discrete Cosine Transform (DCT), the perceptual hash sequence of the image was obtained by comparing the relationship between the coefficients of DCT and the mean of the coefficients. Finally, the encrypted medical image retrieval was achieved by comparing the normalized correlation coefficients between the perceived hash sequences. Compared with the hash algorithm based on Non-negative Matrix Factorization (NMF), the proposed algorithm improves the retrieval accuracy by nearly 40% under Gaussian noise, which is not changed obviously under the JPEG compression attack, median filter attack, scaling attack and ripple distortion attack. Experimental results show that the proposed algorithm has strong robustness against geometric attack and conventional attack, as well as reduce the time complexity of image encryption.