发布时间: 2018-05-16
DOI: 10.11834/jig.170467
2018 | Volume 23 | Number 5




expand article info 丁凯孟1,2,3, 杨晓梅2, 苏守宝3, 刘岳明2
1. 金陵科技学院网络与通信工程学院, 南京 211169;
2. 中国科学院地理科学与资源研究所资源与环境信息系统国家重点实验室, 北京 100101;
3. 金陵科技学院数据科学与智慧软件江苏省重点实验室, 南京 211169


目的 多光谱遥感影像的完整性、真实性等安全问题逐步受到人们的关注,但是,传统认证技术更多地关注数据载体的认证,其不能满足多光谱遥感影像的认证需求。针对多光谱遥感影像的数据特点,提出一种融合波段感知特征的多光谱遥感影像感知哈希认证算法。方法 首先,采用隐形格网划分将多光谱影像的各个波段划分成不同的区域;然后,采用离散小波变换对各波段相同地理位置的格网单元进行分解,并分别采用不同的融合规则对小波变换后的不同分量进行融合;最后,通过Canny算子与奇异值分解提取融合结果的感知特征,再对提取的感知特征进行归一化,最终生成影像的感知哈希序列。多光谱影像的认证过程通过精确匹配感知哈希序列来实现。结果 本文算法采用Landsat TM影像和高分二号卫星的融合影像数据为实验对象,从摘要性、可区分行、鲁棒性、算法运行效率以及安全性等方面进行测试与分析。结果表明,该算法只需要32字节的认证信息就能够实现多光谱遥感影像的认证,摘要性有了较大提高,且算法运行效率提高约1倍;同时,该算法可以有效检测影像的恶意篡改,并对无损压缩和LSB水印嵌入保持近乎100%的鲁棒性。结论 本文算法克服了现有技术在摘要性、算法运行效率等方面不足,而且有较好的可区分性、鲁棒性,能够用于多光谱遥感影像的完整性认证,尤其适合对摘要性要求较高的环境。


多光谱遥感影像; 感知哈希; 波段融合; 完整性认证; 边缘特征; 离散小波变换

Perceptual hash algorithm based on band feature fusion for multispectral remote sensing image authentication
expand article info Ding Kaimeng1,2,3, Yang Xiaomei2, Su Shoubao3, Liu Yueming2
1. School of Networks and Tele-Communications Engineering, Jinling Institute of Technology, Nanjing 211169, China;
2. State Key Laboratory of Resource and Environment Information System, Institute of Geographic Sciences and Natural Resources Research, CAS, Beijing 100101, China;
3. Jiangsu Key Laboratory of Data Science and Smart Software, Nanjing 211169, China
Supported by: National Natural Science Foundation of China (41501431, 61375121);Natural Science Foundation of Jiangsu Province, China(BK200170116)


Objective Given the rapid growth of remote sensing techniques, multispectral remote sensing images exhibit increasing potential for more applications. However, a multispectral image can be easily tampered with or forged during transmission and processing because of the widespread use of sophisticated image editing tools, which threatens the integrity of its content and reduces its value. Therefore, ensuring the content credibility and authenticity of multispectral imaged is a major issue before such images are used. However, existing authentication technologies cannot meet requirements because they are sensitive to each bit of the input data. Perceptual hashing, also known as robustness hash, is able to solve the problems of multispectral image content authentication. Perceptual hash has been developed as a frontier research topic in the field of digital media content security and can be applied to image content authentication, image retrieval, image registration, and digital watermarking. Similar to cryptographic hash functions, perceptual hash compresses the representation of the perceptual features of an image to generate a compact feature vector called perceptual hash value, which is a short summary of the perceptual content of an image. Although perceptual hash for the authentication of normal images has been extensively investigated, research on perceptual hashing for multispectral image authentication is limited. The bands of the MS remote sensing image obtain information from the visible and near-infrared spectra of reflected light, which have clear physical meanings. A multispectral image is composed of a set of monochrome images of the same scene, whereas a normal color image is composed of only three monochrome images and grayscale image that has only one channel. The existing perceptual hash algorithm essentially does not take this into account and cannot perceive the content of each band. In light of the data characteristics of the multispectral remote sensing image, a perceptual hash algorithm based on band feature fusion for multispectral image authentication is proposed in this study. Method The algorithm consists of four main stages, i.e., preprocessing, band fusion, feature extraction, and hash value generation. First, taking the large amount of data of multispectral image into consideration, the bands of the multispectral image are partitioned into grids. Given that the tamper location capability is built on the resolution of grid division, the choice of the grid division resolution presents the trade-off between cost and tamper location capability. Second, the grids with the same geographic location are decomposed and fused by two-level discrete wavelet transform, in which different fusion rules are used by low-frequency, intermediate-frequency, and high-frequency components to keep as many fringe features as possible. For intermediate-frequency components, the fusion rule of "maximum first" is selected; for low-frequency and high-frequency components, the adaptive weighted fusion is selected. This stage is intended for encoding the grids of the source bands into a single grid that contains the best aspects of the original grids, which could be suitable for hashing computation. Third, the edge features of the fusion result are extracted based on the Canny operator to construct the edge feature matrix. Given that the hash value has to be as compact and robust as possible to preserve content, the significant singular values are selected as the perceptual features of the fusion result after singular value decomposition on the matrix. Then, the selected singular value is normalized by the hash function to generate the perceptual hash value of the multispectral image. The number of singular values selected depends on the robustness requirement of the algorithm, and the security of the perceptual hash value depends on the selected hash function. The authentication process is implemented through a precise comparison between reconstructed and original perceptual hash values, and the tamper location can be determined if necessary. Result The experiments indicate that the proposed algorithm can achieve content integrity authentication for multispectral remote sensing images with only 32 bytes of authentication information and has good sensitivity to detect local detailed tampering of the multispectral image, such as removing an object, appending an object, and changing an object. The comparison of the hash values of each grid can be used to identify the tamper location and the corresponding geographic region, and the location granularity depends on the resolution of the grid division. By contrast, the proposed algorithm has approximately 100% robustness to lossless compression, has the least significant bit watermark embedding, and has relatively good robustness to lossy compression. In addition, the computational efficiency of the proposed algorithm is doubled that of the existing algorithm. The robustness of the algorithm can be adjusted by setting the number of selected singular values of the feature matrix. Conclusion The experiments and discussions show that the proposed algorithm is sensitive to malicious tampering and is robust to content-preserving operations on multispectral images, whereas the hash value is relatively compact and the computational efficiency is relatively high. The algorithm can meet the requirement of integrity authentication for multispectral remote sensing image.

Key words

multi-spectral remote sensing images; perceptual hashing; band fusion; integrity authentication; edge features; discrete wavelet transform (DWT)

0 引言


现有的认证技术主要包括密码学方法、数字水印技术以及感知哈希技术等。其中,密码学方法主要基于密码学Hash函数和数字签名生成数据的认证信息,进而实现数据认证。但是,密码学方法针对数据进行二进制级别的认证,只要数据发生一个比特的变化,都视为数据发生篡改(即“雪崩效应”)。这种敏感性比较适合文本数据,并不适合图像(包括遥感影像)、视频、音频等多媒体数据。比如,多光谱影像经过无损数据压缩后,其有效信息并没有改变,不影响数据的使用。数字水印技术主要通过嵌入到数据的认证信息来进行数据认证,当数据内容受到怀疑时,提取并检测嵌入的认证信息是否发生变化,继而鉴别数据的真伪。但是,数字水印技术会对原始数据进行或多或少的修改,而且,数字水印技术主要是利用了水印本身的性质,并不能检测数据的有效内容是否发生篡改。多光谱遥感影像认证本质上关注的是其承载的有效内容信息是否完整,而不是信息的载体,密码学技术与数字水印技术并不能很好地解决多光谱遥感影像的认证问题。感知哈希(Perceptual Hashing)则能够克服传统认证技术的不足,能够更好地实现多光谱遥感影像的数据认证。



1 算法理论基础

1.1 感知哈希



根据多媒体对象的不同,感知哈希大致可以分为图像感知哈希[3]、视频感知哈希[4-6]以及音频感知哈希[7]等。对于图像感知哈希算法,常用的特征提取方式包括:基于图像统计的方法[3]、基于图像粗略特征表示的方法[8]、基于关系的方法[9]、基于特征点的方法10]、基于矩阵分解的方法[11]、基于边缘特征的方法[12]等。近些年,针对不同的图像应用领域,一些新颖的图像感知哈希算法相继被提出。Chen等人[13]提出一种基于压缩感知(compressive sensing)的感知哈希算法,通过稀疏矩阵进行特征降维,但该算法主要面向视觉跟踪(visual tracking)领域,不能直接用于遥感图像认证。Qin等人[14]提出一种基于显著结构特征(salient structure features)的感知哈希算法,将原始图像划分为互不重叠的区域后,再通过离散余弦变换(DCT)提取鲁棒特征,该算法能够用于图像认证与图像检索。Yang等人[15]采用一种基于低密度奇偶校验码的分布式信源编码来压缩特征,继而生成感知哈希序列,具有较好的摘要性。Tang等人[16]提出一种基于颜色向量角(color vector angle)和Canny算子的感知哈希算法,对旋转变换等具有较好的鲁棒性。Yan等人[17]提出一种基于自适应局部特征的感知哈希算法,通过自适应特征点检测与平稳小波变换(stationary wavelet transform)提取图像特征,能够有效检测恶意篡改,并具有较高鲁棒性。Cui等人[18]提出一种面向3D图像感知哈希算法,通过三级双树复小波变换(three-level dualtree complex wavelet transform)提取图像特征并生产哈希序列。陈飞等人[19]针对多标签图像的检索问题,提出一种基于卷积神经网络和目标提取的哈希生成方法,提高了多标签图像检索的性能。


1.2 多光谱遥感影像的数据特征




1.3 基于DWT的波段特征融合




2 算法流程描述

本文算法的总体流程如图 1所示。首先,采用隐形格网划分将多光谱遥感影像的各个波段划分成不同的区域;然后,基于两级离散小波变换(DWT)对相同地理位置的格网单元进行分解,分别采用不同的融合规则对小波变换后的低频、中频和高频分量进行融合;接下来,采用Canny算子与奇异值分解(SVD)提取融合结果的感知特征,并通过Hash函数对提取的感知特征进行归一化,最终生成多光谱影像的感知哈希序列。多光谱遥感影像的认证过程通过比较影像的感知哈希序列是否发生变化来实现。

图 1 融合波段特征的多光谱遥感影像感知哈希认证算法流程
Fig. 1 The flow chart of the proposed perceptual hash algorithm for multi-spectral remote sensing image authentication

2.1 多光谱遥感影像的预处理

对多光谱遥感影像的各个波段进行$ W \times H$的隐形格网划分,将各个波段划分成大小相等且互不重叠的格网单元。划分后的格网单元记为$ G_{wh}^k$,其中,$ k$表示格网所在的波段,$ w$$ h$标识格网在相应波段的位置。格网划分的粒度,也就是$ W$$ H$的选择,应当综合考虑计算效率、影像实际大小、篡改定位粒度等多方面因素。

2.2 面向特征提取的波段融合


2.2.1 低频信息的融合规则

本文算法采用自适应加权的方法对各个波段的低频信息进行融合,其中,加权系数由格网单元的信息熵决定。令$ \mathit{\alpha }_k$表示格网单元的低频分量在融合过程中的加权系数,$ E_{wh}^k$表示相应格网单元的信息熵,$ N$表示波段数,则

$ {{\mathit{\alpha }}_{\mathit{k}}}\rm{=}\frac{\mathit{E}_{^{\mathit{wh}}}^{\mathit{k}}}{\sum\limits_{\mathit{i}=1}^{\mathit{N}}{\mathit{E}_{^{\mathit{wh}}}^{\mathit{i}}}} $ (1)

令低频分量融合的结果记为$ FLL_{wh}$,那么$ FLL_{wh}$的每一个像素为

$ \mathit{FL}{{\mathit{L}}_{\mathit{wh}}}\left( \mathit{i}\rm{, }\mathit{j} \right)\rm{=}\sum\limits_{\mathit{k}=1}^{\mathit{N}}{{{\mathit{\alpha }}_{\mathit{k}}}\mathit{FLL}_{^{\mathit{wh}}}^{\mathit{k}}\left( \mathit{i}\rm{, }\mathit{j} \right)} $ (2)

式中,$ FLL_{wh}^k (i, j)$表示第$ k$波段中相应格网单元的像素值。

2.2.2 中频信息的融合规则

由于中频信息相比高频信息具有一定的抗噪性,而且,绝对值较大的中频系数对应的是较为明显的边缘特征,所以,为了尽可能地保留明显的边缘特征,本文算法采用“选择极大值”的规则来融合各个波段的频信息。这里,令水平中频分量、垂直中频分量、对角中频分量的融合结果分别表示为$ FHL_{wh}$$ FLH_{wh}$$ FHH_{wh}$,则$ FHL_{wh}$的每一个像素为

$ \begin{align} & \mathit{FH}{{\mathit{L}}_{\mathit{wh}}}\left( \mathit{i}\rm{, }\mathit{j} \right)\rm{=max } \{ \rm{ }\mathit{FHL}_{\mathit{wh}}^{1}\left( \mathit{i}\rm{, }\mathit{j} \right)\rm{, } \\ & \ \ \ \ \mathit{FHL}_{\mathit{wh}}^{2}\left( \mathit{i}\rm{, }\mathit{j} \right)\rm{, }\cdots \rm{, }\mathit{FHL}_{^{\mathit{wh}}}^{\mathit{N}}\left( \mathit{i}\rm{, }\mathit{j} \right)\rm{ } \} \rm{ } \\ \end{align} $ (3)

图 2所示为根据上述过程进行融合的某TM影像的水平中频分量的融合结果,其中,图 2(a)(c)分别为第3波段In$ R_3$、第4波段In$ R_4$、第7波段In$ R_7$的水平中频分量,图 2(d)为融合结果。对比可知,融合结果较好地保留了各波段中较明显的边缘特征。

图 2 水平中频分量融合实例
Fig. 2 Fusion instance of IF((a) IF of the grid in In$ R_3$; (b) IF of the grid in In$ R_4$; (c) IF of the grid in In$ R_7$; (d) fusion result)


2.2.3 高频信息的融合规则



2.3 特征提取与处理

对融合后的格网单元${\mathit{\boldsymbol{F}}}_{w, h}$进行特征提取,进一步生成影像的感知哈希序列。本文采用Canny算子提取融合结果的边缘特征,并通过奇异值分解提取前$ K$个奇异值作为感知特征,具体过程为:

1) 采用插值算法将融合后的格网单元${\mathit{\boldsymbol{F}}}_{w, h}$的分辨率变为$ m$×$ m$像素(本文实验中,$ m$=64)。此举不仅为了降低计算复杂度,还能降低噪声影响。

2) 过Canny算子提取格网单元$F_{w, h}$的边缘特征,并进行0-1序列化,得到的边缘特征矩阵记为$ \mathit{\boldsymbol{ME}}$

3) 为了消除感知哈希鲁棒性与敏感性之间的矛盾,对$ \mathit{\boldsymbol{ME}}$进行SVD分解,提取前$ K$个奇异值作为格网单元的感知特征。其中,$ K$的取值由算法鲁棒性要求和矩阵大小而定,本文实验中选取矩阵前1/4的奇异值。

2.4 感知哈希序列的生成

顾及遥感影像的高精度要求,本文采用Hash函数(这里以MD5为例)对提取的奇异值进行归一化,记为$ H_{w, h}$。串联所有格网单元的$ H_{w, h}$,并通过Hash函数(本文以SHA-256为例,实际情况中也可以选用强度更高的Hash函数,如SHA-512)进行压缩,得到的摘要

$ \mathit{SH}\rm{=}\ \rm{SHA-256(}{{\mathit{H}}_{\rm{0, 0}}}\left\| {{\mathit{H}}_{\rm{0, 1}}} \right.\left\| \cdots \right\|{{\mathit{H}}_{\mathit{W}\rm{, }\mathit{H}}}\rm{)} $ (4)

另一方面,为了满足篡改定位的需求,同时顾及算法的摘要性,本文取$ H_{w, h}$的前32位作为该格网单元的哈希序列,记为$ PH_{w, h}$$ SH$$ PH_{w, h}$进行串联,就得到多光谱影像的感知哈希序列

$ \mathit{PH}\rm{=}\mathit{SH}\left\| \mathit{P}{{\mathit{H}}_{\rm{0, 0}}} \right.\left\| \mathit{P}{{\mathit{H}}_{\rm{0, 1}}} \right.\left\| \cdots \right\|\mathit{P}{{\mathit{H}}_{\mathit{W}\rm{, }\mathit{H}}} $ (5)

式中,只需要$ SH$就可以实现多光谱影像的完整性认证,而$ PH_{w, h}$能够对篡改进行定位。

2.5 多光谱遥感影像的认证过程

本文算法基于“精确匹配”的方式来比较待认证多光谱影像的哈希序列$ SH$′与原始影像的哈希序列$ SH$之间的差异来实现影像的完整性认证。如果两者相同,说明待认证影像的内容没有发生明显变化;反之,说明待认证多光谱影像遭到了某种篡改,即

$ \left\{ \begin{align} & \rm{影像内容完整}\ \ \ \ \ \ \ \mathit{SH}\rm{=}\mathit{S{H}'} \\ & \rm{影像被篡改}\ \ \ \ \ \ \ \ \ \ \mathit{SH}\ne \mathit{S{H}'} \\ \end{align} \right. $ (6)

进一步比较$ PH_{w, h}$可以将篡改定位到具体的地理区域。

3 实验结果与分析

分别以Landsat TM影像(7个波段)和高分二号卫星的融合影像数据(3个波段)为例对提出的算法进行测试与分析:图 3图 4图 5所示分别为选取的南京地区的Landsat TM影像(记为影像A)、某山区小镇的Landsat TM影像(记为影像B)、高分二号卫星的银川地区融合影像(记为影像C),各波段大小分别为1 536×1 024像素、512×512像素、4 096×4 096像素。

图 3 测试影像A的不同波段
Fig. 3 Bands of the TM image A ((a) In$ R_1$; (b) In$ R_4$; (c) In$ R_7$)
图 4 测试影像B的不同波段
Fig. 4 Bands of the TM image B((a) In$ R_1$; (b) In$ R_4$; (c) In$ R_5$; (d) In$ R_6$)
图 5 测试影像C(高分二号卫星的融合影像)
Fig. 5 Test image C ((a) original image; (b) blue band)

关于格网划分的粒度,对影像B和C分别进行4×4和16×16的格网划分;对影像A则分别进行6×4和12×8的两种粒度的格网划分,以测试不同格网划分粒度对算法的影响。实验的硬件平台为:2.4 GHz主频的双核CPU,内存4 GB;软件开发平台为Visual Studio 2013,编程语言采用C++,并基于GDAL库函数实现部分功能。

3.1 可区分性

多光谱影像的认证过程必须能够检测出波段的局部篡改。影像的某波段被篡改之后,相应格网单元的哈希序列和影像的感知哈希序列$ SH$都将发生变化,这样,就能够检测出恶意篡改。进一步比较格网单元的哈希序列,能够对篡改进行定位(定位粒度为相应的地理格网区域)。

图 6图 8所示为实验影像A、影像B与影像C的篡改实例,分别是删除地物、增加地物、变换地物等篡改操作。表 1所示为实验影像A篡改前后(图 6(a)(d)所示)、影像B篡改前后(图 6(b)(e)所示)与影像C篡改前后(图 6(c)(f)所示)的哈希序列变换情况。

图 6 篡改测试实例1(删除地物)
Fig. 6 Tamper test (removing the object) ((a) In$ R_1$ of the original image A; (b) In$ R_1$ of the original image B; (c) original grid of image C; (d) In$ R_1$ of the tampered image A; (e) In$ R_1$ of the tampered image B; (f) tampered grid of image C)
图 7 测试实例2(增加地物)
Fig. 7 Tamper test (appending the object) ((a) In$ R_3$ of the original image A; (b) In$ R_5$ of the original image B; (c) original grid of image C; (d) In$ R_3$ of the tampered image A; (e) In$ R_5$ of the tampered image B; (f) tampered grid of image C)
图 8 篡改测试实例3(变换地物)
Fig. 8 Tamper test (changing the object)((a) In$ R_7$ of the original image A; (b) In$ R_7$ of the original image B; (c) original grid of image C; (d) In$ R_7$ of the tampered image A; (e) In$ R_7$ of the tampered image B; (f) tampered grid of image C)

表 1 篡改前后的哈希序列对比
Table 1 The comparison of the hash values

原始哈希序列 影像篡改后的哈希序列
$ PH_{w, h}$ $ SH$ $ PH_{w, h}$ $ SH$
实验影像A 56a2
实验影像B f421
实验影像C a409



3.2 鲁棒性

鲁棒性是感知哈希与密码学认证技术的最大区别。数据压缩和不可见数字水印嵌入是典型的不改变多光谱影像内容的操作,因此以数据压缩和水印嵌入为例进行鲁棒性测试,其中,数据压缩分别采用无损压缩和有损压缩,水印嵌入以最低有效位(LSB)为例,表 2为测试结果。这里采用感知哈希序列未发生变化的格网单元所占百分比来描述算法鲁棒性。

表 2 鲁棒性测试结果
Table 2 The result of robustness test

影像 无损压缩 水印嵌入 JPEG压缩
影像A 6×4划分 100 100 87.5
影像A 12×8划分 100 100 86.4
影像B 100 100 93.7
影像C 100 100 88.3

表 2可知,本文算法对多光谱影像的无损压缩、LSB水印嵌入等操作具有较好的鲁棒性,对有损压缩也能保持一定鲁棒性。形成对比的是,密码学认证方法均视上述操作为非法篡改,不能实现较好的认证。


3.3 算法运行效率与安全性分析

下面测试本文算法的计算效率,并与文献[20]的算法进行比较(软硬件平台相同,且均采用C++编程语言),结果见表 3

表 3 计算效率
Table 3 The computational efficiency

算法 影像A
影像B 影像C
文献[20] 16.41 30.32 11.21 34.94
本文 9.76 12.26 6.98 16.41



4 结论



  • [1] Du B, Zhang L F, Zhang L P, et al. Discriminant manifold learning approach for hyperspectral image dimension reduction[J]. Acta Photonica Sinica, 2013, 42(3): 320–325. [杜博, 张乐飞, 张良培, 等. 高光谱图像降维的判别流形学习方法[J]. 光子学报, 2013, 42(3): 320–325. ] [DOI:10.3788/gzxb20134203.0320]
  • [2] Niu X M, Jiao Y H. An overview of perceptual hashing[J]. Acta Electronica Sinica, 2008, 36(7): 1405–1411. [牛夏牧, 焦玉华. 感知哈希综述[J]. 电子学报, 2008, 36(7): 1405–1411. ] [DOI:10.3321/j.issn:0372-2112.2008.07.029]
  • [3] Xiang S J, Kim H J, Huang J W. Histogram-based image hashing scheme robust against geometric deformations[C]//Proceedings of the 9th workshop on Multimedia & Security. Dallas, Texas: ACM, 2007: 121-128. [DOI:10.1145/1288869.1288886]
  • [4] Sandeep R, Sharma S, Thakur M, et al. Perceptual video hashing based on Tucker decomposition with application to indexing and retrieval of near-identical videos[J]. Multimedia Tools and Applications, 2016, 75(13): 7779–7797. [DOI:10.1007/s11042-015-2695-1]
  • [5] Ouyang J, Gao J H, Wen Z K, et al. Video perceptual hashing fuse computational model of human visual system[J]. Journal of Image and Graphics, 2011, 16(10): 1883–1889. [欧阳杰, 高金花, 文振焜, 等. 融合HVS计算模型的视频感知哈希算法研究[J]. 中国图象图形学报, 2011, 16(10): 1883–1889. ] [DOI:10.11834/jig.20111005]
  • [6] Zhu Y Y, Wen Z K, Du Y H, et al. Video forgery detection and multi-granularity location based on video perceptual hashing[J]. Journal of Image and Graphics, 2013, 18(8): 924–932. [朱映映, 文振焜, 杜以华, 等. 基于视频感知哈希的视频篡改检测与多粒度定位[J]. 中国图象图形学报, 2013, 18(8): 924–932. ] [DOI:10.11834/jig.20130806]
  • [7] Zhang Q Y, Xing P F, Liu Y W, et al. Research on security transmission of perceptual hash values based on ECC and digital watermarking[J]. International Journal of Security and Its Applications, 2015, 9(3): 255–266. [DOI:10.14257/ijsia.2015.9.3.19]
  • [8] Saad S M. Design of a robust and secure digital signature scheme for image authentication over wireless channels[J]. IET Information Security, 2009, 3(1): 1–8. [DOI:10.1049/iet-ifs:20070112]
  • [9] Lu C S, Liao H Y M. Structural digital signature for image authentication:an incidental distortion resistant scheme[J]. IEEE Transactions on Multimedia, 2003, 5(2): 161–173. [DOI:10.1109/TMM.2003.811621]
  • [10] Liu Z Q, Li Q, Liu J R, et al. SIFT based image hashing algorithm[J]. Chinese Journal of Scientific Instrument, 2011, 32(9): 2024–2028. [刘兆庆, 李琼, 刘景瑞, 等. 一种基于SIFT的图像哈希算法[J]. 仪器仪表学报, 2011, 32(9): 2024–2028. ] [DOI:10.19650/j.cnki.cjsi.2011.09.017]
  • [11] Xiang S J, Yang J Q. NMF-based image hashing algorithm using restricted random blocking[J]. Journal of Electronics & Information Technology, 2011, 33(2): 337–341. [项世军, 杨建权. 基于约束随机分块的NMF图像哈希算法[J]. 电子与信息学报, 2011, 33(2): 337–341. ] [DOI:10.3724/SP.J.1146.2010.00212]
  • [12] Ding K M, Zhu C Q. Perceptual hash algorithm for integrity authentication of remote sensing image[J]. Journal of Southeast University:Natural Science Edition, 2014, 44(4): 723–727. [丁凯孟, 朱长青. 一种用于遥感影像完整性认证的感知哈希算法[J]. 东南大学学报:自然科学版, 2014, 44(4): 723–727. ] [DOI:10.3969/j.issn.1001-0505.2014.04.008]
  • [13] Chen L, Li Z, Yang J F. Compressive perceptual hashing tracking[J]. Neurocomputing, 2017, 239: 69–80. [DOI:10.1016/j.neucom.2017.02.004]
  • [14] Qin C, Chen X Q, Dong J, et al. Perceptual image hashing with selective sampling for salient structure features[J]. Displays, 2016, 45: 26–37. [DOI:10.1016/j.displa.2016.09.003]
  • [15] Yang Y C, Zhou J W, Duan F P, et al. Wave atom transform based image hashing using distributed source coding[J]. Journal of Information Security and Applications, 2016, 31: 75–82. [DOI:10.1016/j.jisa.2016.09.001]
  • [16] Tang Z J, Huang L Y, Zhang X Q, et al. Robust image hashing based on color vector angle and canny operator[J]. AEU-International Journal of Electronics and Communications, 2016, 70(6): 833–841. [DOI:10.1016/j.aeue.2016.03.010]
  • [17] Yan C P, Pun C M, Yuan X C. Multi-scale image hashing using adaptive local feature extraction for robust tampering detection[J]. Signal Processing, 2016, 121: 1–16. [DOI:10.1016/j.sigpro.2015.10.027]
  • [18] Cui C, Mao H K, Niu X M, et al. A novel hashing scheme for Depth-image-based-rendering 3D images[J]. Neurocomputing, 2016, 191: 1–11. [DOI:10.1016/j.neucom.2016.01.028]
  • [19] Chen F, Lyu S H, Li J, et al. Multi-label image retrieval by hashing with object proposal[J]. Journal of Image and Graphics, 2017, 22(2): 232–240. [陈飞, 吕绍和, 李军, 等. 目标提取与哈希机制的多标签图像检索[J]. 中国图象图形学报, 2017, 22(2): 232–240. ] [DOI:10.11834/jig.20170211]
  • [20] Ding K M, Zhu C Q, Su S B, et al. Perceptual hash algorithm for integrity authentication of multispectral remote sensing images[J]. Optics and Precision Engineering, 2015, 23(10): 676–683. [丁凯孟, 朱长青, 苏守宝, 等. 用于多光谱影像完整性认证的感知哈希算法[J]. 光学精密工程, 2015, 23(10): 676–683. ] [DOI:10.3788/OPE.20152313.0677]
  • [21] Liu F Z, Zhang J X, Lin Z J, et al. Measurement of graylevel and texture information in multispectral images[J]. Geomatics and Information Science of Wuhan University, 2016, 41(3): 415–420. [刘凤珠, 张景雄, 林宗坚, 等. 多光谱遥感影像的灰度与纹理信息测度方法[J]. 武汉大学学报:信息科学版, 2016, 41(3): 415–420. ] [DOI:10.13203/j.whugis20140329]
  • [22] Zhang L B, Chen J, Qiu B C. Region of interest extraction in remote sensing images by saliency analysis with the normal directional lifting wavelet transform[J]. Neurocomputing, 2016, 179: 186–201. [DOI:10.1016/j.neucom.2015.11.093]
  • [23] Song M X, Guo P. A combinatorial optimization method for remote sensing image fusion with Contourlet and HSI transform[J]. Journal of Computer-Aided Design & Computer Graphics, 2012, 24(1): 83–88. [宋梦馨, 郭平. 结合Contourlet和HSI变换的组合优化遥感图像融合方法[J]. 计算机辅助设计与图形学学报, 2012, 24(1): 83–88. ] [DOI:10.3969/j.issn.1003-9775.2012.01.014]
  • [24] Yang Y, Tong S, Huang S Y. Image fusion based on fast discrete Curvelet transform[J]. Journal of Image and Graphics, 2015, 20(2): 219–228. [杨勇, 童松, 黄淑英. 快速离散Curvelet变换域的图像融合[J]. 中国图象图形学报, 2015, 20(2): 219–228. ] [DOI:10.11834/jig.20150208]
  • [25] Cheng J, Liu H J, Liu T, et al. Remote sensing image fusion via wavelet transform and sparse representation[J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2015, 104: 158–173. [DOI:10.1016/j.isprsjprs.2015.02.015]