Local Feature Vectors Research Articles

Locally lateral manifolds of normalised Gabor features for face recognition

Due to inherent characteristics of multiscale and orientation, normalised Gabor features have been successfully used in face recognition. Various previous works have showcased the strength and feasibility of this approach, especially on its robustness against local variations. However, the projected features are numerous and substantial in dimension, which is largely due to the convolution of multiscale and orientation of wavelets. Such features, when used in practical face recognition, would require relatively lengthy classification process, particularly when it involves computationally extensive local classifier or experts, such as ensembles of local cosine similarity (ELCS) classifier. The authors address this issue by simultaneously reducing the size of Gabor features laterally and locally using a manifold learning method called locally linear embedding (LLE). This method is thus denoted as locally lateral normalised local Gabor feature vector with LLE (LGFV/LN/LLE). Results on several publicly available face datasets reveal the superiority of the authors’ approach in terms of improvements in feature compression of LGFV features by up to a reduction of 95% of total dimensionality while increasing the average classification accuracy by 26%. Altogether, the authors show that their LGFV/LN/LLE augmented by ELCS classifiers delivers equivalent result when compared against the state‐of‐the‐art.

Fast Image Index for Database Management Engines

Abstract Large-scale image repositories are challenging to perform queries based on the content of the images. The paper proposes a novel, nested-dictionary data structure for indexing image local features. The method transforms image local feature vectors into two-level hashes and builds an index of the content of the images in the database. The algorithm can be used in database management systems. We implemented it with an example image descriptor and deployed in a relational database. We performed the experiments on two image large benchmark datasets.

Fingerprint image quality assessment based on BP neural network with hierarchical clustering

Fingerprint image quality assessment is important because the good performance of the minutiae-based matching algorithm is heavily dependent on fingerprint images with high quality. Many efforts have been made in existing methods, but most methods either use full fingerprint images or use local areas and involve subjective judgments. Unlike previous methods, the proposed method considers both local and global assessments. Local feature vectors are extracted from the fingerprint image block for hierarchical clustering, and the results are used as target outputs of the back-propagation (BP) neural network without any subjective judgments. Global feature vectors based on the local quality assessment results are used for hierarchical clustering and fed into the BP neural network that calculates the overall error rate of genuine and imposter errors to achieve global quality assessment. Furthermore, the minutiae quality assessment method is also proposed and incorporated into the minutiae-based matching algorithm. The experimental results based on the FVC2002 and FVC2004 databases show that the proposed methods can effectively assess the quality of fingerprint images and ensure the overall improvement of matching performance.

Image Classification Model Using Visual Bag of Semantic Words

In the image classification field, the visual bag of words (BoW) has two drawbacks. One is low classification accuracy because a visual BoW is typically extracted from local low-level visual feature vectors via key points, without considering the high-level semantics of an image. The other is excessive time consumption because the size of the vocabulary is very large, especially for images with explicit backgrounds and object content. To solve these two problems, we propose a novel image classification model based on a visual bag of semantic words (BoSW), which includes an automatic segmentation algorithm based on graph cuts to extract major semantic regions and a semantic annotation algorithm based on support vector machine to label the regions with a visual semantic vocabulary. The proposed BoSW model refines image semantics by introducing user conceptions for extracting semantic vocabularies and reducing the size of the vocabulary. Experimental results demonstrate the superiority of the proposed algorithm through comparisons with state-of-the-art methods on benchmark datasets.

Panoramic Image Mosaics via Distributed Systems Using Color Moments and Local Wavelet-Features

In this paper, an efficient method based on color moments and local wavelet-features is proposed for panoramic image mosaics. Color moments are used to efficiently represent physical quantities of objects in an image. Wavelet can describe a wide variety of image characteristics and is a key component for image-related applications. Therefore, during the feature-extraction stage for panoramic image mosaics, we exploit color moments and wavelet-subband statistics to construct local feature vectors for image-patch representation. With a distributed system of local area network, the proposed mosaics system can achieve an average 295 FPS. Experimental results show that the local wavelet-features are able to produce plausible and satisfactory panoramic images.

A novel method for SIFT features matching based on feature dimension matching degree

We proposes a method for fast matching SIFT feature points based on SIFT feature descriptor vector element matching. First, we discretize each dimensional feature element into an array address based on a fixed threshold value and store the corresponding feature point labels in an address. If the same dimensional feature element of the descriptor vector has the same discrete value, their feature point labels may fall into the same address. Secondly, we search the mapping address of the feature descriptor vector element to obtain the matching state of the corresponding dimensions of the feature descriptor vector, thus obtaining the number of dimensions matching between feature points and feature dimension matching degree. Then we use the feature dimension matching degree to obtain the suspect matching feature points. Finally we use the Euclidean distance to eliminate the mismatching feature points to obtain accurate matching feature point pairs. The method is essentially a high-dimensional feature vector matching method based on local feature vector element matching. Experimental results show that the new algorithm can guarantee the number of matching SIFT feature points and their matching accuracy and that its running time is similar to that of HKMT, RKDT and LSH algorithms

A Content-Based Image Retrieval Method By Exploiting Cluster Shapes

Content-Based Image Retrieval (CBIR) is an automatic process of retrieving images that are the most similar to a query image based on their visual content such as colour and texture features. However, CBIR faces the technical challenge known as the semantic gap between high level conceptual meaning and the low-level image based features. This paper presents a new method that addresses the semantic gap issue by exploiting cluster shapes. The method first extracts local colours and textures using Discrete Cosine Transform (DCT) coefficients. The Expectation-Maximization Gaussian Mixture Model (EM/GMM) clustering algorithm is then applied to the local feature vectors to obtain clusters of various shapes. To compare dissimilarity between two images, the method uses a dissimilarity measure based on the principle of Kullback-Leibler divergence to compare pair-wise dissimilarity of cluster shapes. The paper further investigates two respective scenarios when the number of clusters is fixed and adaptively determined according to cluster quality. Experiments are conducted on publicly available WANG and Caltech6 databases. The results demonstrate that the proposed retrieval mechanism based on cluster shapes increases the image discrimination, and when the number of clusters is fixed to a large number, the precision of image retrieval is better than that when the relatively small number of clusters is adaptively determined.

Collaborative representation based face classification exploiting block weighted LBP and analysis dictionary learning

Traditional collaborative representation based classification (CRC) method usually faces the challenge of data uncertainty hence results in poor performance, especially in the presence of appearance variations in pose, expression and illumination. To overcome this issue, this paper presents a CRC-based face classification method by jointly using block weighted LBP and analysis dictionary learning. To this end, we first design a block weighted LBP histogram algorithm to form a set of local histogram-based feature vectors instead of using raw images. By this means we are able to effectively decrease data redundancy and uncertainty derived from image noises and appearance variations. Second, we adopt an analysis dictionary learning model as the projection transform to construct an analysis subspace, in which a new sample is characterized with the improved sparsity of its reconstruction coefficient vector. The crucial role of the analysis dictionary learning method in CRC is revealed by its capacity of the collaborative representation in an analytic coefficient space. Extensive experimental results conducted on a set of well-known face databases demonstrate the merits of the proposed method.

A Node Selection Paradigm for Crowdsourcing Service Based on Region Feature in Crowd Sensing

Crowd sensing is a human-centered sensing model. Through the cooperation of multiple nodes, an entire sensing task is completed. To improve the efficiency of sensing missions, a cost-effective set of service nodes, which is easy to fit in performing different tasks, is needed. In this paper, we propose a low-cost service node selection method based on region features, which builds on the relationship between task requirements and geographical locations. The method uses Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm to cluster service nodes and calculate the center point of each cluster. The area then is divided into regions according to rules of Voronoi diagrams. Local feature vectors are constructed according to the historical records in each divided region. When a particular sensing task arrives, Analytic Hierarchy Process (AHP) is used to match the feature vector of each region to mission requirements to get a certain number of service nodes satisfying the characteristics. To get a lower cost output, a revised Greedy Algorithm is designed to filter the exported service nodes to get the required low-cost service nodes. Experimental results suggest that the proposed method shows promise in improving service node selection accuracy and the timeliness of finishing tasks.

Machine learning enhanced global optimization by clustering local environments to enable bundled atomic energies.

We show how to speed up global optimization of molecular structures using machine learning methods. To represent the molecular structures, we introduce the auto-bag feature vector that combines (i) a local feature vector for each atom, (ii) an unsupervised clustering of such feature vectors for many atoms across several structures, and (iii) a count for a given structure of how many times each cluster is represented. During subsequent global optimization searches, accumulated structure-energy relations of relaxed structural candidates are used to assign local energies to each atom using supervised learning. Specifically, the local energies follow from assigning energies to each cluster of local feature vectors and demanding the sum of local energies to amount to the structural energies in the least squares sense. The usefulness of the method is demonstrated in basin hopping searches for 19-atom structures described by single- or double-well Lennard-Jones type potentials and for 24-atom carbon structures described by density functional theory. In all cases, utilizing the local energy information derived on-the-fly enhances the rate at which the global minimum energy structure is found.

Ensemble Decision for Spam Detection Using Term Space Partition Approach.

This paper proposes an ensemble decision approach which combines global and local features of e-mails together to detect spam effectively. In the proposed method, a special feature construction method named term space partition (TSP) is utilized to divide the whole term space into several subspaces and adopt different feature construction strategies on each of them, respectively. This method can make each term play a distinct and important role when conducting detection. This method is utilized and extended by introducing the sliding window technique to extract local features from e-mails. The global classifier and local classifiers are constructed on a global feature vector set and local feature vector sets, respectively, and together make the ensemble decision by adopting the voting technique. The principles of the TSP-based approach and mechanism of the ensemble decision method are presented in detail. Five different and standard benchmark corpora are applied to experiments for performance evaluation of this proposed method. Comprehensive experimental results show that the proposed method brings significant performance improvement and better robustness on the basis of the TSP-based approach. In addition, the proposed method outperforms the current prevalent and state-of-the-art approaches, especially when a comprehensive consideration of performance, efficiency, and robustness is taken. This endows it with flexible capability and adaptivity in the real-world applications.

Feature covariance matrix-based dynamic hand gesture recognition

Over the past 2 decades, vision-based dynamic hand gesture recognition (HGR) has made significant progresses and been widely adopted in many practical applications. Although the advent of RGB-D cameras and deep learning-based methods provides more feasible solutions for HGR, it is still very challenging to satisfy the requirements of both high efficiency and accuracy for real-world HGR systems. In this paper, we propose a novel method using the feature covariance matrix for effective and efficient dynamic HGR. We extract a set of local feature vectors that represent local motion patterns to construct the feature covariance matrix efficiently, which also provides a compact representation of a dynamic hand gesture. By tracking hand keypoints in three successive frames and calculating their motion features, our method can be extended to both 2D dynamic HGR and 3D dynamic HGR. To evaluate the effectiveness of the proposed framework, we perform extensive experiments on three publicly available datasets (one 2D dataset and two 3D datasets). The experimental results demonstrate the effectiveness of our proposed method.

Context-Dependent Random Walk Graph Kernels and Tree Pattern Graph Matching Kernels with Applications to Action Recognition.

Graphs are effective tools for modeling complex data. Setting out from two basic substructures, random walks and trees, we propose a new family of context-dependent random walk graph kernels and a new family of tree pattern graph matching kernels. In our context-dependent graph kernels, context information is incorporated into primary random walk groups. A multiple kernel learning algorithm with a proposed l1,2-norm regularization is applied to combine context-dependent graph kernels of different orders. This improves the similarity measurement between graphs. In our tree-pattern graph matching kernel, a quadratic optimization with a sparse constraint is proposed to select the correctly matched tree-pattern groups. This augments the discriminative power of the tree-pattern graph matching. We apply the proposed kernels to human action recognition, where each action is represented by two graphs which record the spatiotemporal relations between local feature vectors. Experimental comparisons with state-of-the-art algorithms on several benchmark datasets demonstrate the effectiveness of the proposed kernels for recognizing human actions. It is shown that our kernel based on tree-pattern groups, which have more complex structures and exploit more local topologies of graphs than random walks, yields more accurate results but requires more runtime than the context-dependent walk graph kernel.

A new computer-aided detection approach based on analysis of local and global mammographic feature asymmetry.

This study aims to develop and test a new computer-aided detection (CAD) approach and scheme, assessing the likelihood of a subject harboring breast abnormalities. The proposed scheme is based on the analysis of both local and global bilateral mammographic feature asymmetries. The level of local or global asymmetry is assessed by analyzing mammographic features extracted from the bilaterally matched regions of interest (ROIs), or from the entire breast, respectively. The selected local and global feature vectors are combined and classified using a maximum likelihood obtained from a naïve Bayes classifier. This scheme was evaluated using a leave-one-case-out cross-validation method that was applied to 243 subjects from mini-MIAS and INbreast databases. In addition, the result is compared with a conventional unilateral (or single) image-based CAD scheme. Using a case-based evaluation approach and an area under curve (AUC) of the receiver operating characteristic (ROC) as a performance index, the new scheme yielded AUC=0.79±0.07, an 8.2% increase compared with AUC=0.73±0.08 obtained using the unilateral image-based CAD scheme. This work demonstrates that applying bilateral asymmetry analysis increases the discriminatory power of CAD schemes while optimizing the likelihood assessment of breast abnormalities presence. Therefore, the proposed CAD approach provides the radiologist with beneficial supplementary information and can indicate high-risk cases.

A quick survey of Artificial Neural Network based face classification algorithms

Face recognition has received significant attention in the recent past and numerous applications have deployed. Face recognition systems produce pleasing performance under controlled scenarios, and they produce a poor performance with real-world scenarios. High recognition rate and less training time are the two primary requirements in face recognition system. In this survey, various neural network based classification techniques for face recognition application are considered, such as Artificial Neural Network, Adaptive CNN (ACNN), convolutional neural network, Extreme Learning Machine and probabilistic neural network. The accuracy is calculated based on local, global and hybrid feature vector. Finally, paper summaries the algorithms in-terms of some crucial parameters and highlights the suitable algorithm for various applications.

Automatic segmentation and classification of olive fruits batches based on discrete wavelet transform and visual perceptual texture features

The quality of olive fruit and its virgin olive oil is a main concern for consumers and fruit industrial companies. The effectiveness and fast detection of olive’s skin defects is the most decisive factor in determining its quality. It is necessary to design and implement image processing tools for segmentation and correct classification of the different fresh incoming olive batches. In this paper, we propose a new automatic image segmentation algorithm, based on discrete wavelets transform. The aim of the segmentation algorithm is to discriminate between olives and the background with the challenge of irregular and dispersive lesion borders, low contrast, artifacts in the olive fruit and variety of colors within the interest region. The second part of our work proposes a scheme for olive fruit classification. The classifier first identifies the olive fruit color and then, based upon discrete wavelets transform and Tamura statistical texture features, the healthy olive fruit is distinguished from the damaged one. The new texture feature vector is, then, compared with the robust Local Binary Pattern feature vector. The simplicity of our segmentation and classification algorithms makes them appropriate for designing a productive and profitable computer vision machine.

Feature Representation Using Deep Autoencoder for Lung Nodule Image Classification

This paper focuses on the problem of lung nodule image classification, which plays a key role in lung cancer early diagnosis. In this work, we propose a novel model for lung nodule image feature representation that incorporates both local and global characters. First, lung nodule images are divided into local patches with Superpixel. Then these patches are transformed into fixed‐length local feature vectors using unsupervised deep autoencoder (DAE). The visual vocabulary is constructed based on the local features and bag of visual words (BOVW) is used to describe the global feature representation of lung nodule image. Finally, softmax algorithm is employed for lung nodule type classification, which can assemble the whole training process as an end‐to‐end mode. Comprehensive evaluations are conducted on the widely used public available ELCAP lung image database. Experimental results with regard to different parameter setting, data augmentation, model sparsity, classifier algorithms, and model ensemble validate the effectiveness of our proposed approach.

An integrated framework for posture recognition

Postures can be identified and then classified from the video sequences using scale invariant keys as classification features and the results of classification can be used in various fields like surveillance, medical diagnosis and training purpose. In this paper frames are extracted from the given video files and transformed into a large collection of local feature vectors using scale invariant feature transform (SIFT), each of which is not affected by image translation, scaling and rotation, and to some extent invariant to illumination changes and affine or 3D projection. Features are grouped using k-means clustering in which each posture belongs to the cluster with the nearest mean. Multiclass support vector machine i.e., directed acyclic graph (DAGSVM) then assigns labels to the centres obtained from clustering. AdaBoost is incorporated to boost the performance accuracy of the classifier. Dataset used in this study is Bharatnatyam video dataset. The posture classification model is also shown to outperform state-of-art classification systems on videos as classification accuracy achieved using this frame work is 89%.

An advanced fingerprint matching using minutiae-based indirect local features

Biometric systems examine the uniqueness of an individual based on physical and behavioral characteristics. Among the known traits, fingerprint is the most significant biometric trait due to its ease of use and high accuracy. However, the efficiency of the fingerprint matching technique depends on the feature vector it uses. The ideal feature vector should be invariant to several common transformations, which usually a fingerprint capturing system is subjected to. Current work focuses to achieve such an invariance by extracting the features based on the spatial relationship among minutiae points. We propose a minutiae point based 4-dimensional local feature vector, which simultaneously satisfies six desirable feature vector properties. This feature vector definition helps us to deal with problem of missing and spurious minutiae and thus enables us to design a robust authentication system. We have substantiated the efficacy of the proposed approach with the help of a number of fingerprint instances available in FVC and NIST databases.

A new approach for image detection based on refined Bag of Words algorithm

This paper presents an efficient image retrieval approach based on enhanced Bag of Words (BOW). By keeping eyes on low classification efficiency and accuracy of existing image retrieval system based on traditional BOW, a new classification method combined features of scale invariant feature transform (SIFT) and GIST principle (CGSF) is proposed. This method use SIFT algorithm to extract the local feature vector and GIST algorithm to extract the global feature vector for images. The feature fusion scheme is applied to combine local with global features through weight which can enhance the classification accuracy with efficiency. At the end, VLFeat linear support vector machine (SVM) classifier is used to classify the visual dictionary and Osirix medical computed tomography (CT) image dataset is selected for experimental verification. Three experiments are carried out based on factors that contribute to performance and accuracy (dictionary size and λ co-efficient). The experimental results prove that our proposed technique is much robust and accurate than traditional algorithms in image retrieval. Furthermore, our method is matched with literature in classification accuracy and the outcomes indicate the noticeable benefits in medical, internet of things and many other domains.