Sharif Digital Repository

New methods based on local spatio-temporal features have exhibited significant performance in action recognition. In these methods, feature selection plays an important role to achieve a superior performance. Actions are represented by local spatio-temporal features extracted from action videos. Action representations are then classified by applying a classifier (such as k-nearest neighbor or SVM). In this paper, we have proposed two feature weighting methods to better discriminate similar actions. We have proposed a definition of feature discrimination power to be used in the feature selection process. Our proposed weighting schemes have greatly improved the final categorization accuracy on... 

Sport video shot classification is a basic step in the sport video processing. For many purposes such as event detection and summarization, shot classification is needed for content filtering. In this paper, we present a new method for soccer video shot classification. At first, in-field and out-of-field frames are separated. In in-field frames three features based on number of connected components and shirt color percent in vertical and horizontal strips are extracted. The features are all new and showed excellent discrimination in the feature space. These features are given to SVM for classifying long, medium and close-up shots. One of the advantages of our method is that, close-ups can be... 

In this paper we propose a new automatic method for discriminating original and tampered images based on 'JPEG ghost detection' method, which is a subset of format-based image forensics approaches. The inconsistency of quality factors indicates that the photo is a composite one created from at least two different cameras and therefore it is a manipulated photo. Our classification algorithm first extracts the ghost border. Then the image is classified as original or tampered groups by thresholding a distance in feature space. © 2016 IEEE  

In this paper, we address the problem of data description using a Bayesian framework. The goal of data description is to draw a boundary around objects of a certain class of interest to discriminate that class from the rest of the feature space. Data description is also known as one-class learning and has a wide range of applications. The proposed approach uses a Bayesian framework to precisely compute the class boundary and therefore can utilize domain information in form of prior knowledge in the framework. It can also operate in the kernel space and therefore recognize arbitrary boundary shapes. Moreover, the proposed method can utilize unlabeled data in order to improve accuracy of... 

In this paper, we propose a mapping from low level feature space to the semantic space drawn by the users through relevance feedback to enhance the performance of current content based image retrieval (CBIR) systems. The proposed approach makes a rule base for its inference and configures it using the feedbacks gathered from users during the life cycle of the system. Each rule makes a hypercube (HC) in the feature space corresponding to a semantic concept in the semantic space. Both short and long term strategies are taken to improve the accuracy of the system in response to each feedback of the user and gradually bridge the semantic gap. A scoring paradigm is designed to determine the... 

To enable a relevance feedback paradigm to evolve itself by users' feedback, a reinforcement learning method is proposed. The feature space of the medical images is partitioned into positive and negative hypercubes by the system. Each hypercube constitutes an individual in a genetic algorithm infrastructure. The rules take recombination and mutation operators to make new rules for better exploring the feature space. The effectiveness of the rules is checked by a scoring method by which the ineffective rules will be omitted gradually and the effective ones survive. Our experiments on a set of 10,004 images from the IRMA database show that the proposed approach can better describe the semantic... 

In this paper, we propose a novel density-based fuzzy clustering algorithm based on Active Learning Method (ALM), which is a methodology of soft computing inspired by some hypotheses claiming that human brain interprets information in pattern-like images rather than numerical quantities. The proposed clustering algorithm, Fuzzy Unsupervised Active Learning Method (FUALM), is performed in two main phases. First, each data point spreads in the feature space just like an ink drop that spreads on a sheet of paper. As a result of this process, densely connected ink patterns are formed that represent clusters. In the second phase, a fuzzifying process is applied in order to summarize the effects... 

In this paper, we propose a novel density-based fuzzy clustering algorithm based on Active Learning Method (ALM), which is a methodology of soft computing inspired by some hypotheses claiming that human brain interprets information in pattern-like images rather than numerical quantities. The proposed clustering algorithm, Fuzzy Unsupervised Active Learning Method (FUALM), is performed in two main phases. First, each data point spreads in the feature space just like an ink drop that spreads on a sheet of paper. As a result of this process, densely connected ink patterns are formed that represent clusters. In the second phase, a fuzzifying process is applied in order to summarize the effects... 

Shape analysis plays an important role in many medical imaging studies. One of the recent shape analysis methods uses the Laplace Beltrami operator which is also used in this paper for hippocampal shape comparison. We proposed a feature vector which consists of size measures and shape descriptors based on Laplace Beltrami eigenvalues and eigenfunctions. The aforementioned feature space is utilised for automatic differentiating normal subjects from epileptic patients as well as distinguishing epileptic patients with left temporal lobe epilepsy (LTLE) from patients with right temporal lobe epilepsy (RTLE). Achieved results are diagnostic accuracy of 93% with 95% sensitivity and lateralization... 

Ambient Intelligence is the future of computing where devices predict what users need and help them carry out their everyday life activities easier. To make this prediction possible these environments should be aware of the context. Activity recognition is one of the most complex problems in context-aware environments. In this paper we propose a layered Dynamic Bayesian Network (DBN) to recognize activities in an oral presentation. The layered architecture gives us the opportunity to recognize complex activities using the classification results of sensory data in the first layer regardless of the physical environment. Our model is event-driven meaning the classification takes place only when... 

In this study, clustering based method for activation detection in functional magnetic resonance imaging (fMRI) is employed. Moreover, some features are obtained by fitting two models namely FIR filter and Gamma function, to hemodynamic response function (HRF). After applying clustering methods (that require number of clusters as an input) to feature space, our simulations show that number of clusters can affect activation detection significantly. Therefore a newly proposed clustering algorithm namely evolving neural gas (ENG) that gives optimal number of clusters is exploited. In addition to ENG, the result of four clustering algorithms namely k-means, fuzzy C-means, neural gas, and clara... 

Background and Objectives: Melanoma is a dangerous kind of skin disease with a high death rate, and its prevalence has increased rapidly in recent years. Diagnosis of melanoma in a primary phase can be helpful for its cure. Due to costs for dermatology, we need an automatic system to diagnose melanoma through lesion images. Methods: Here, we propose a sparse representation based method for segmentation and classification of lesion images. The main idea of our framework is based on a kernel sparse representation, which produces discriminative sparse codes to represent features in a high-dimensional feature space. Our novel formulation for discriminative kernel sparse coding jointly learns a... 

In the image segmentation field, traditional techniques do not completely meet the segmentation challenges mostly posed by inherently fuzzy images. Fuzzy connectedness and fuzzy clustering are considered as two well-known techniques for introducing fuzzy concepts to the problem of image segmentation. Fuzzy connectedness-based Segmentation methods consider spatial relation of image pixels by "hanging togetherness" a notion based on intensity homogeneity. But, they do not inherently utilize feature information of image pixels. On the other hand, as the segmentation domain of fuzzy clustering-based methods is the feature space they do not consider spatial relations among image pixels. Recently,... 

In this paper, we present new adaptive linear discriminant analysis (LDA) algorithm and apply them for adaptive facial feature extraction. Adaptive nature of the proposed algorithm is advantageous for real world applications in which one confronts with a sequence of data such as online face recognition and mobile robotics. Application of the new algorithm on feature extraction from facial image sequences is given in three steps: i) adaptive image preprocessing, ii) adaptive dimension reduction and iii) adaptive LDA feature estimation. Steps 1 and 2 are done simultaneously and outputs of stage 2 are used as a sequence of inputs for stage3. The proposed system was tested on Yale and PIE face...