Robust Image Hashing for Tamper Detection Using Non-Negative Matrix Factorization

Zhenjun Tang,Shuozhong Wang,Xinpeng Zhang,Weimin Wei,Shengjun Su 대한전자공학회 2008 JUCT : Journal of Ubiquitous Convergence Technolog Vol.2 No.1

The invariance relation existing in the non-negative matrix factorization (NMF) is used for constructing robust image hashes in this work. The image is first re-scaled to a fixed size. Low-pass filtering is performed on the luminance component of the re-sized image to produce a normalized matrix. Entries in the normalized matrix are pseudo-randomly re-arranged under the control of a secret key to generate a secondary image. Non-negative matrix factorization is then performed on the secondary image. As the relation between most pairs of adjacent entries in the NMF's coefficient matrix is basically invariant to ordinary image processing, a coarse quantization scheme is devised to compress the extracted features contained in the coefficient matrix. The obtained binary elements are used to form the image hash after being scrambled based on another key. Similarity between hashes is measured by the Hamming distance. Experimental results show that the proposed scheme is robust against perceptually acceptable modifications to the image such as Gaussian filtering, moderate noise contamination, JPEG compression, re-scaling, and watermark embedding. Hashes of different images have very low collision probability. Tampering to local image areas can be detected by comparing the Hamming distance with a predetermined threshold, indicating the usefulness of the technique in digital forensics.

이미지 기반 SNS의 스토리텔링 구조 연구

김유나(Juna Kim) 인문콘텐츠학회 2016 인문콘텐츠 Vol.0 No.41

SNS는 개인의 미시적인 일상을 공유할 수 있는 스토리텔링 매체이다. 본 연구에서는 SNS가 텍스트 기반에서 이미지 기반의 형식으로 변화하고 있다는 점에 주목하고, 이미지 기반 SNS에 나타난 스토리텔링 구조를 밝히고자 한다. 이때 이미지와 해시태그(#)가 빈번하게 결합한다는 점에 주목하고 그 결합 양상을 분석한다. 이미지 기반 SNS에서 이미지는 다의적 의미로 해석될 수 있는 비결정적 사건이며, 해시태그는 이미지를 해석하는 기표가 된다. 특히 해시태그는 계열적 발산 운동을 통해 이미지에 대한 서사적 맥락을 생성하고, 이는 다시 수렴 운동을 통해 하나의 서사적 의미로 귀결한다. 이미지 기반 SNS는 이와 같은 이중 운동을 통해 서사적 구조를 획득한다. 본 연구에서는 이미지와 해시태그가 끊임없이 상호작용을 하면서 서사적 의미를 생성해낸다는 점에서 이미지 기반 SNS의 매체적 함의를 고찰한다. SNS is a storytelling medium in which users can share their personal daily lives. This study takes note that text-based SNS is turning into image-based one, and tries to reveal the storytelling structure of image-based SNS. Specifically this study pays attention to the fact that images and hash tags(#) are frequently binding together and analyzes the pattern. In the image-based SNS, images function as non-deterministic events that can be interpreted in several ways. Hash tags function as interpretative signifier of the image. In particular hash tag is creating a narrative context for the images through paradigmatic divergent movement, which then led to a convergence of narrative meaning through syntagmatic convergent movement. Image-based SNS acquires the narrative structure through this dual movement. This study thus examines the implications of image-based SNS in that it is constantly generating narrative meanings from interactions between images and hash tags.

히스토그램 기반의 강인한 계층적 GLOCAL 해쉬 생성 방법

최용수(Yong Soo Choi),김형중(Hyoung Joong Kim),이달호(Dal Ho Lee) 大韓電子工學會 2011 電子工學會論文誌-CI (Computer and Information) Vol.48 No.1

최근 들어, 웹 응용의 하나로 이미지를 통합 관리하는 이미지 거래소(Image Stock), 이미지 도서관(Image Library)과 같은 응용들이 많이 만들어 지고 있다. 이미지의 등록, 관리, 검색에는 주로 이미지 해쉬라는 기술이 구분자(Identifier)로서 쓰이며 해쉬의 분별력을 높이기 위한 연구들이 많이 진행되어지고 있다. 본 논문에서는 계층적 히스토그램을 이용한 GLOCAL(Global to Local) 이미지 해쉬 생성 방법을 제안하였다. 많은 연구들이 이미지 처리 및 기하학적 공격에 강한 히스토그램 기반의 이미지 해쉬 기법들을 제안하였으며 제안된 논문에서는 GLOCAL 해쉬 생성과 가중치(Weighting Factor)를 적용하여 해쉬의 안정성을 높이는데 기여하였다. GLOCAL 해쉬 생성 방법에 의해 기존의 알고리즘들은 좀더 풍부한 길이의 이미지 해쉬를 생성하였다. 즉, 이미지 해쉬의 근본 목적인 Identification과 Discrimination 이라는 두 가지 목적을 잘 달성하였으며 그 결과는 통계학적 가설 검정(Statistical Hypothesis Testing)을 통해 기존의 알고리즘과 비교하였으며 대부분의 공격종류에 대해 제안된 알고리즘이 향상된 성능을 보여줌을 확인하였다. Recently, Web applications, such as Stock Image and Image Library, are developed to provide the integrated management for user's images. Image hash techniques are used for the image registration, management and retrieval as the identifier and many researches have been performed to raise the hash performance. This paper proposes GLOCAL image hashing method utilizing the hierarchical histogram which based on histogram bin population method. So far, many researches have proven that image hashing techniques based on histogram are robust image processing and geometrical attack. We modified existing image hashing method developed by our research team. The main idea is that it makes more fluent hash string if we have histogram bin of specific length as shown in the body of paper. Finally, we can raise the magnitude of hash string within same context or feature and strengthen the robustness of hash.

Robust Image Hashing for Tamper Detection Using Non-Negative Matrix Factorization

Tang, Zhenjun,Wang, Shuozhong,Zhang, Xinpeng,Wei, Weimin,Su, Shengjun The Institute of Electronics and Information Engin 2008 JUCT : Journal of Ubiquitous Convergence Technolog Vol.2 No.1

The invariance relation existing in the non-negative matrix factorization (NMF) is used for constructing robust image hashes in this work. The image is first re-scaled to a fixed size. Low-pass filtering is performed on the luminance component of the re-sized image to produce a normalized matrix. Entries in the normalized matrix are pseudo-randomly re-arranged under the control of a secret key to generate a secondary image. Non-negative matrix factorization is then performed on the secondary image. As the relation between most pairs of adjacent entries in the NMF's coefficient matrix is basically invariant to ordinary image processing, a coarse quantization scheme is devised to compress the extracted features contained in the coefficient matrix. The obtained binary elements are used to form the image hash after being scrambled based on another key. Similarity between hashes is measured by the Hamming distance. Experimental results show that the proposed scheme is robust against perceptually acceptable modifications to the image such as Gaussian filtering, moderate noise contamination, JPEG compression, re-scaling, and watermark embedding. Hashes of different images have very low collision probability. Tampering to local image areas can be detected by comparing the Hamming distance with a predetermined threshold, indicating the usefulness of the technique in digital forensics.

엔트로피 연산자를 이용한 영상 해싱 기반 인식자

박제호,Park, Je-Ho 한국반도체디스플레이기술학회 2021 반도체디스플레이기술학회지 Vol.20 No.3

The desire for a technology that can mechanically acquire 2D images starting with the manual method of drawing has been making possible a wide range of modern image-based technologies and applications over a period. Moreover, this trend of the utilization of image-related technology as well as image-based information is likely to continue. Naturally, as like other technology areas, the function that humans produce and utilize by using images needs to be automated by using computing-based technologies. Surprisingly, technology using images in the future will be able to discover knowledge that humans have never known before through the information-related process that enables new perception, far beyond the scope of use that human has used before. Regarding this trend, the manipulation and configuration of massively distributed image database system is strongly demanded. In this paper, we discuss image identifier production methods based on the utilization of the image hashing technique which especially puts emphasis over an entropy operator.

Concentric Circle-Based Image Signature for Near-Duplicate Detection in Large Databases

Ayoung Cho,Won-Keun Yang,Dong-Seok Jeong,Weon-Geun Oh 한국전자통신연구원 2010 ETRI Journal Vol.32 No.6

Many applications dealing with image management need a technique for removing duplicate images or for grouping related (near-duplicate) images in a database. This paper proposes a concentric circle-based image signature which makes it possible to detect near-duplicates rapidly and accurately. An image is partitioned by radius and angle levels from the center of the image. Feature values are calculated using the average or variation between the partitioned sub-regions. The feature values distributed in sequence are formed into an image signature by hash generation. The hashing facilitates storage space reduction and fast matching. The performance was evaluated through discriminability and robustness tests. Using these tests, the particularity among the different images and the invariability among the modified images are verified, respectively. In addition, we also measured the discriminability and robustness by the distribution analysis of the hashed bits. The proposed method is robust to various modifications, as shown by its average detection rate of 98.99%. The experimental results showed that the proposed method is suitable for near-duplicate detection in large databases.

Perceptual Bound-Based Asymmetric Image Hash Matching Method

서진수 한국멀티미디어학회 2017 멀티미디어학회논문지 Vol.20 No.10

Image hashing has been successfully applied for the problems associated with the protection of intellectual property, management of large database and indexation of content. For a reliable hashing system, improving hash matching accuracy is crucial. In order to improve the hash matching performance, we propose an asymmetric hash matching method using the psychovisual threshold, which is the maximum amount of distortion that still allows the human visual system to identity an image. A performance evaluation over sets of image distortions shows that the proposed asymmetric matching method effectively improves the hash matching performance as compared with the conventional Hamming distance.

Hashing via Efficient Addictive Kernel for Logistics Image Classification

Xiao-jun Liu,Qiu-ling Li,Bin Zhang,Jun-yi Li 보안공학연구지원센터 2016 International Journal of Database Theory and Appli Vol.9 No.2

In this paper, fast image search with efficient additive kernels and kernel locality-sensitive hashing has been proposed. As to hold the kernel functions, recent work has probed methods to create locality-sensitive hashing , which guarantee our approach’s linear time, however existing methods still do not solve the problem of locality-sensitive hashing (LSH) and indirectly sacrifice the loss in accuracy of search results in order to allow fast queries. To improve the search accuracy, we show how to apply explicit feature maps into the homogeneous kernels, which help in feature transformation and combine it with kernel locality-sensitive hashing. We prove our method on several large datasets, and illustrate that it improve the accuracy relative to commonly used methods and make the task of object classification, content-based retrieval more fast and accurate.

Perceptual Bound-Based Asymmetric Image Hash Matching Method

Seo, Jiin Soo Korea Multimedia Society 2017 멀티미디어학회논문지 Vol.20 No.10

Image hashing has been successfully applied for the problems associated with the protection of intellectual property, management of large database and indexation of content. For a reliable hashing system, improving hash matching accuracy is crucial. In order to improve the hash matching performance, we propose an asymmetric hash matching method using the psychovisual threshold, which is the maximum amount of distortion that still allows the human visual system to identity an image. A performance evaluation over sets of image distortions shows that the proposed asymmetric matching method effectively improves the hash matching performance as compared with the conventional Hamming distance.

Robust image hashing using SIFT feature points and DWT approximation coefficients

Lokanadham Naidu Vadlamudi,Rama Prasad V. Vaddella,Vasumathi Devara 한국통신학회 2018 ICT Express Vol.4 No.3

This study proposes a robust hashing method using scale-invariant feature transform (SIFT) features points and discrete wavelet transform (DWT) approximation coefficients for image authentication. Initially, the invariant feature points are computed using SIFT from the component of color image. Next, n distinct SIFT feature points are utilized to extract image content from the component. Then, DWT is applied to extracted content in order to compute approximation coefficients. Finally, the approximation coefficients are normalized to form a binary hash. Experimental results show that the proposed method is robust to various content-preserving operations such as compression, scaling, filtering, additive noise, brightness, and contrast adjustment. In addition, the performance of the proposed method is compared to existing methods using a receiver operating characteristics curve. The comparison results show that the proposed method performs better than the existing methods.