• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제12권9호
• /
• pp.4412-4428
• /
• 2018

The purpose of this work is to solve the problem of representing an entire video using Convolutional Neural Network (CNN) features for human action recognition. Recently, due to insufficient GPU memory, it has been difficult to take the whole video as the input of the CNN for end-to-end learning. A typical method is to use sampled video frames as inputs and corresponding labels as supervision. One major issue of this popular approach is that the local samples may not contain the information indicated by the global labels and sufficient motion information. To address this issue, we propose a binary hashing method to enhance the local feature extractors. First, we extract the local features and aggregate them into global features using maximum/minimum pooling. Second, we use the binary hashing method to capture the motion features. Finally, we concatenate the hashing features with global features using different normalization methods to train the classifier. Experimental results on the JHMDB and MPII-Cooking datasets show that, for these new local features, binary hashing mapping on the sparsely sampled features led to significant performance improvements.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제12권8호
• /
• pp.3790-3803
• /
• 2018

Content-based image retrieval is an approach used to query images based on their semantics. Semantic based retrieval has its application in all fields including medicine, space, computing etc. Semantically generated binary hash codes can improve content-based image retrieval. These semantic labels / binary hash codes can be generated from unlabeled data using convolutional autoencoders. Proposed approach uses semi-supervised deep hashing with semantic learning and binary code generation by minimizing the objective function. Convolutional autoencoders are basis to extract semantic features due to its property of image generation from low level semantic representations. These representations of images are more effective than simple feature extraction and can preserve better semantic information. Proposed activation and loss functions helped to minimize classification error and produce better hash codes. Most widely used datasets have been used for verification of this approach that outperforms the existing methods.

• 정보처리학회논문지:소프트웨어 및 데이터공학
• /
• 제3권6호
• /
• pp.215-218
• /
• 2014

해슁(hashing)을 기반으로 한 근사 최근접 이웃 탐색(approximate nearest neighbors search, ANN search) 방법에서는 데이터 샘플들을 k-비트 이진 코드로 변환하는 해쉬 함수들을 이용함으로써 근접 이웃 탐색이 이진변환 공간에서 이루어지게 된다. 본 논문에서는 PCA 기반 군집화 방법인 Principal Direction Divisive Partitioning(PDDP)를 이용한 해슁 방법을 제안한다. PDDP는 가장 큰 분산을 가지는 클러스터를 선택하여 그 클러스터의 첫 번째 주성분 방향을 이용하여 두 개의 클러스터로 분할하는 과정을 반복적으로 시행하는 군집화 방법이다. 제안하는 해슁 방법에서는 PDDP에서 분할을 위해 사용하는 주성분방향을 바이너리 코딩을 위한 사영벡터로서 사용한다. 실험결과는 제안하는 방법이 다른 해슁 방법들과 비교하여 경쟁력 있는 방법임을 입증한다.

• 전자공학회논문지CI
• /
• 제47권6호
• /
• pp.75-85
• /
• 2010

본 논문에서는 3D 콘텐츠 인증을 위한 객체별 특징 벡터 기반 강인한 3D 모델 해싱을 제안한다. 제안한 3D 모델 해싱에서는 다양한 객체들로 구성된 3D 모델에서 높은 면적을 가지는 특징 객체내의 꼭지점 거리들을 그룹화한다. 그리고 각 그룹들을 치환한 다음, 그룹 계수, 랜덤 변수 키와 이진화 과정에 의하여 최종 해쉬를 생성한다. 이 때 해쉬의 강인성은 객체 그룹별 꼭지점 거리 분포를 그룹 계수에 의하여 향상되고, 해쉬의 유일성은 그룹 계수를 치환 키 및 랜덤변수 키 기반의 이진화 과정에 의하여 향상된다. 실험 결과로부터 제안한 해싱이 다양한 메쉬 공격 및 기하학 공격에 대한 해쉬의 강인성과 유일성을 확인하였다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제9권12호
• /
• pp.5058-5072
• /
• 2015

Quaternions have been commonly employed in color image processing, but when the existing pure quaternion representation for color images is used in perceptual hashing, it would degrade the robustness performance since it is sensitive to image manipulations. To improve the robustness in color image perceptual hashing, in this paper a full quaternion representation for color images is proposed by introducing the local image luminance variances. Based on this new representation, a novel Full Quaternion Discrete Cosine Transform (FQDCT)-based hashing is proposed, in which the Quaternion Discrete Cosine Transform (QDCT) is applied to the pseudo-randomly selected regions of the novel full quaternion image to construct two feature matrices. A new hash value in binary is generated from these two matrices. Our experimental results have validated the robustness improvement brought by the proposed full quaternion representation and demonstrated that better performance can be achieved in the proposed FQDCT-based hashing than that in other notable quaternion-based hashing schemes in terms of robustness and discriminability.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제9권7호
• /
• pp.2599-2613
• /
• 2015

Multi-index hashing (MIH) is the state-of-the-art method for indexing binary codes, as it di-vides long codes into substrings and builds multiple hash tables. However, MIH is based on the dataset codes uniform distribution assumption, and will lose efficiency in dealing with non-uniformly distributed codes. Besides, there are lots of results sharing the same Hamming distance to a query, which makes the distance measure ambiguous. In this paper, we propose a data-oriented multi-index hashing method (DOMIH). We first compute the covariance ma-trix of bits and learn adaptive projection vector for each binary substring. Instead of using substrings as direct indices into hash tables, we project them with corresponding projection vectors to generate new indices. With adaptive projection, the indices in each hash table are near uniformly distributed. Then with covariance matrix, we propose a ranking method for the binary codes. By assigning different bit-level weights to different bits, the returned bina-ry codes are ranked at a finer-grained binary code level. Experiments conducted on reference large scale datasets show that compared to MIH the time performance of DOMIH can be improved by 36.9%-87.4%, and the search accuracy can be improved by 22.2%. To pinpoint the potential of DOMIH, we further use near-duplicate image retrieval as examples to show the applications and the good performance of our method.

• 한국음향학회지
• /
• 제32권5호
• /
• pp.415-422
• /
• 2013

본 논문에서는 경첩 손실 함수를 최소화를 통해서 강인한 이진 오디오 핑거프린팅 방법을 제안하였다. 특히 제안된 방법에서 오디오 핑거프린트는 이진값을 가지므로 핑거프린트 DB 크기를 줄여줄 수 있는 장점이 있다. 일반적으로 특징을 이진화하는 과정에서 핑거프린트의 강인성, 식별성 등 성능의 손실이 불가피하므로 손실을 최소화하는 것이 필요하다. 본 논문에서는 핑거프린팅에서 두 오디오 클립 간의 유사도가 경첩 함수 형태로 주어지는 것에 착안하여 경첩 손실을 최소화하는 방법으로 특징을 이진화하여 핑거프린트를 구하는 방법을 제안한다. 유도된 경첩 손실 함수는 최소 손실 해싱 기법을 통해서 최소화 하였다. 수 천곡 규모의 오디오에 대해서 다양한 변환들에 대한 인식 성능을 실험하였으며, 제안된 경첩 손실 함수 최소화를 통해서 핑거프린트의 식별성과 강인성이 개선됨을 확인하였다.

• 전자공학회논문지CI
• /
• 제47권1호
• /
• pp.1-14
• /
• 2010

3D 콘텐츠 산업 분야의 급격한 성장과 더불어, 3D 콘텐츠 인증 및 신뢰, 검색을 위한 콘텐츠 해싱 기술이 요구되어지고 있다. 그러나 영상 및 동영상과 같은 2D 콘텐츠 해싱에 비하여 3D 콘텐츠 해싱에 대한 연구가 아직까지 미비하다. 본 논문에서는 키 기반의 3D 표면 계수 분포를 이용한 강인한 3D 메쉬 모델 해싱 기법을 제안한다. 제안한 기법에서는 기본적인 Euclid 기하학 변환에 강인한 3D SSD와 표면 곡률의 평면계 기반의 블록 표면 계수를 특징 벡터로 사용하며, 이를 치환 키 및 랜덤 변수 키에 의하여 최종 이진 해쉬를 생성한다. 실험 결과로부터 제안한 해싱 기법은 다양한 기하학 및 위상학 공격에 강인하며, 모델 및 키별로 해쉬의 유일성을 확인하였다.

• International Journal of Advanced Culture Technology
• /
• 제5권1호
• /
• pp.76-82
• /
• 2017

An image hash, a discriminative and robust summary of an image, should be robust against quality-preserving signal processing steps, while being pairwise independent for perceptually different inputs. In order to improve the hash matching performance, this paper proposes a probabilistic dissimilarity matching. Instead of extracting the binary hash from the query image, we compute the probability that the intermediate hash vector of the query image belongs to each quantization bin, which is referred to as soft quantization binning. The probability is used as a weight in comparing the binary hash of the query with that stored in a database. A performance evaluation over sets of image distortions shows that the proposed probabilistic matching method effectively improves the hash matching performance as compared with the conventional Hamming distance.

• International Journal of Computer Science & Network Security
• /
• 제23권9호
• /
• pp.141-149
• /
• 2023

Repacked mobile apps constitute about 78% of all malware of Android, and it greatly affects the technical ecosystem of Android. Although many methods exist for repacked app detection, most of them suffer from performance issues. In this manuscript, a novel method using the Constant Key Point Selection and Limited Binary Pattern (CKPS: LBP) Feature extraction-based Hashing is proposed for the identification of repacked android applications through the visual similarity, which is a notable feature of repacked applications. The results from the experiment prove that the proposed method can effectively detect the apps that are similar visually even that are even under the double fold content manipulations. From the experimental analysis, it proved that the proposed CKPS: LBP method has a better efficiency of detecting 1354 similar applications from a repository of 95124 applications and also the computational time was 0.91 seconds within which a user could get the decision of whether the app repacked. The overall efficiency of the proposed algorithm is 41% greater than the average of other methods, and the time complexity is found to have been reduced by 31%. The collision probability of the Hashes was 41% better than the average value of the other state of the art methods.