Electronic Letters on Computer Vision and Image Analysis (ELCVIA - Universitat Autònoma de Barcelona)
Not a member yet
343 research outputs found
Sort by
Retinal Image Analysis Oriented to the Clinical Task
Get PDFOphthalmology can profit greatly from the analysis of digital images because they can aid in establishing an early diagnosis even before the first symptoms appear. This dissertation contributes to the digital analysis of such images and the problems that arise along the imaging pipeline of fundus photography, a field that is commonly referred to as retinal image analysis. We have dealt with and proposed solutions to problems that arise in retinal image acquisition and longitudinal monitoring of retinal disease evolution. Specifically, non- uniform illumination compensation[1], poor image quality [2], automated focusing [3], image segmentation [4], change detection [5], space-invariant (SI) [5] and space-variant (SV) [6] blind deconvolution (BD). Digital retinal image analysis can be effective and cost-efficient for disease management, computer-aided-diagnosis, screening and telemedicine and applicable to a variety of disorders such as glaucoma, macular degeneration, and retinopathy [7, 8]
Fuzzy Multilevel Graph Embedding for Recognition, Indexing and Retrieval of Graphic Document Images
Get PDFThis thesis addresses the problem of lack of efficient computational tools for graph based structural pattern recognition approaches and proposes to exploit computational strength of statistical pattern recognition. It has two fold contributions. The first contribution is a new method of explicit graph embedding. The proposed graph embedding method exploits multilevel analysis of graph for extracting graph level information, structural level information and elementary level information from graphs. It embeds this information into a numeric feature vector. The method employs fuzzy overlapping trapezoidal intervals for addressing the noise sensitivity of graph representations and for minimizing the information loss while mapping from continuous graph space to discrete vector space. The method has unsupervised learning abilities and is capable of automatically adapting its parameters to underlying graph dataset. The second contribution is a framework for automatic indexing of graph repositories for graph retrieval and subgraph spotting. This framework exploits explicit graph embedding for representing the cliques of order 2 by numeric feature vectors, together with classification and clustering tools for automatically indexing a graph repository. It does not require a labeled learning set and can be easily deployed to a range of application domains, offering ease of query by example (QBE) and granularity of focused retrieval
Swarm-based Descriptor Combination and its Application for Image Classification
Get PDFIn this paper, we deal with the descriptor combination problem in image classification tasks. This problem refers to the definition of an appropriate combination of image content descriptors that characterize different visual properties, such as color, shape and texture. In this paper, we propose to model the descriptor combination as a swarm-based optimization problem, which finds out the set of parameters that maximizes the classification accuracy of the Optimum-Path Forest (OPF) classifier. In our model, a descriptor is seen as a pair composed of a feature extraction algorithm and a suitable distance function. Our strategy here is to combine distance scores defined by different descriptors, as well as to employ them to weight OPF edges, which connect samples in the feature space. An extensive evaluation of several swarm-based optimization techniques was performed. Experimental results have demonstrated the robustness of the proposed combination approach
A Novel Interest-Point-Based Background Subtraction Algorithm
Get PDFCurrent Back-Ground Subtraction (BGS) algorithms are pixel-based methods. We propose an Interest-Point(IP)-based BGS algorithm applicable in IP-based Computer Vision application. Based on a block-wiseprocessing strategy, the images are divided into blocks of the same size. IPs inside blocks are dealt withtogether as Events. Throughout the frames, the algorithm stores Events of blocks as well as the numbersof their occurrences (Repetition Index (RI)) in a Binary Tree. The RI is used to classify Events into thebackground and foreground. The background Events appear significantly more than a threshold. The otherswith RI value less than the threshold, are classified as the foreground Events. This event classification isused to label IPs of frames into the foreground and background IPs. Experimental results quantitativelyshow that the proposed algorithm delivers a good subtraction rate in comparison with the other BGS ap-proaches. Moreover, it: creates a map of the background usable for further processing; is robust to changesin illumination; and can keep itself updated to changes in the background
Towards an interactive index structuring system for content-based image retrieval in large image databases
Get PDFIn recent years, the expansion of acquisition devices such as digital cameras, the development of storage and transmission techniques and the success of tablet computers facilitate the development of many large image databases as well as the interactions with the users. This thesis [1] deals with the problem of Content-Based Image Retrieval (CBIR) on these huge masses of data. Traditional CBIR systems generally rely on three phases: feature extraction, feature space structuring and retrieval. In this thesis, we are particularly interested in the structuring phase (normally called indexing phase), which plays a very important role in finding information in large databases. This phase aims at organizing the visual feature descriptors of all images into an efficient data structure in order to facilitate, accelerate and improve further retrieval. We assume that the feature extraction phase is completed and the image feature descriptors which are usually low-level features describing the color, shape, texture, etc. of all images are available. Instead of traditional structuring methods, clustering methods which organize image descriptors into groups of similar objects (clusters), without any constraint on the cluster size, are studied. The aim is to obtain an indexed structure more adapted to the retrieval of high dimensional and unbalanced data. Clustering process can be done without prior knowledge (unsupervised clustering) or with a limited amount of prior knowledge (semi-supervised clustering).Due to the “semantic gap” between high-level semantic concepts expressed by the user via the query and the low-level features automatically extracted from the images, the clustering results and therefore the retrieval results are generally different from the wishes of the user. In this thesis, we proposed to involve the user in the clustering phase so that he/she can interact with the system so as to improve the clustering results, and thus improve the performance of the further retrieval. The idea is as follows. Firstly, images are organized into clusters by using an initial clustering. Then, the user visualizes the clustering result and provides feedback to the system in order to guide the re-clustering phase. The system then re-organizes the dataset by using not only the similarity between objects, but also the feedback given by the user in order to reduce the semantic gap. The interactive loop can be iterated until the clustering result satisfies the user. In the case of large database indexing, we assume that the user has no prior knowledge about the image database. Therefore, an unsupervised clustering method is suitable to be used for the initial clustering, when no supervised information is available yet. Then, after receiving the user feedback in each interactive iteration, a semi-supervised clustering can be used for the re-clustering process.Based on a deep study of the state of the art of different unsupervised clustering methods [4] as well as semi- supervised clustering approaches [2, 3], we propose in this thesis a new interactive semi-supervised clustering model [3] involving the user in the clustering phase in order to improve the clustering results. From the formal analysis of different unsupervised clustering methods [4], we chose to experiment some methods which appear to be the most suitable to be used in an incremental context involving the user in the clustering stage. The hierarchical BIRCH unsupervised clustering (Zhang et al., 1996) which gives the best performance from these experiments [4] is chosen to be used as the initial clustering in our model. Then, an interactive loop in which the user provides the feedback to the system and the system re-organizes the image database using the new semi-supervised clustering method proposed in this thesis is iterated until the clustering result satisfies the user. As the user has no prior knowledge about the image database, it is difficult for him/her to label the clusters or the images in the clusters using classes. Therefore, we provide to the user an interactive interface allowing him/her to easily visualize the clustering result and give feedback to the system. Based on the majority of the images displayed for each cluster, the user can specify, by some simple clicks, relevant or non-relevant images for each cluster. The user can also drag and drop images between clusters in order to change the cluster assignment of some images. Then, supervised information is deduced from the user feedback in order to be used for the re-clustering phase using the proposed semi-supervised clustering method. According to our study of the state of the art of different semi-supervised clustering methods, supervised information may consist of class labels for some objects or pairwise constraints (must-link or cannot-link) between objects. The experimental analysis of different semi-supervised clustering methods in the interactive context [2, 3] shows a high performance of the HMRF-kmeans (Basu et al., 2004) which uses pairwise constraints compared with the other methods. Inspired from the HMRF-kmeans method, we proposed a new semi-supervised clustering method [3] for the re-clustering process. Instead of using pairwise constraints between images, our method uses pairwise constraints between the leaf entries (CF entries) of the BIRCH tree as supervised information for guiding the re-organization of the CF entries in the re-clustering phase. As each CF entry groups a set of similar images, pairwise constraints between images can be replaced by a smaller number of pairwise constraints between CF entries, without reducing the quality of supervised information. And therefore, the processing time could be reduced without decreasing the performance. In our model, after receiving user feedback in each interactive iteration, pairwise constraints can be deduced based not only on the user feedback but also on the neighbourhood information. Neighbourhood information groups images according to the willingness of the user to classify them in the same clusters (via user feedback of all interactive iterations). This kind of information helps to maximize the supervised information (pairwise constraints) gained from a same number of user clicks.In order to avoid the subjective dependence of the clustering results on the human user, a software agent simulating the behaviour of the human user for providing feedback to the system is used for the experimental analysis of our system using different image databases of increasing sizes (Wang, PascalVoc2006, Caltech101, Corel30k). Moreover, different strategies for deducing pairwise constraints from user feedback and neighbour- hood information were investigated. Among these strategies, the strategy which keeps only the most “diffi- cult” constraints (must-link between the most distant objects and cannot-links between the closest objects) was shown to give the best trade-off between the performance and the processing time. Furthermore, the experi- mental results show that our model helps to improve the clustering results by involving the user and that our semi-supervised clustering outperforms the HMRF-kmeans, in both performance and processing time. Note that our clustering structure can be used not only for facilitating the further image retrieval, but also for helping the navigation in large image databases. Moreover, in this thesis, we propose a 2D interface for visualizing the group structure of high dimensional image databases
A PGM-based System for Arabic HandwrittenWord Recognition
Get PDFThis paper describes a system for off-line recognition of handwritten Arabic words. It uses simple andeasily extractable features to construct feature vectors for words in the vocabulary. Some of these features are statistical, based on pixel distributions and local pixel configurations. Others are structural, based on the presence of ascenders, descenders and diacritic points. The system is evolved based on horizontal and vertical Hidden Markov Models and Dynamic Bayesian Network. Our strategy consists of looking for various architectures and selecting those which provide the best recognition performance. Experiments on handwritten Arabic words from IFN/ENIT database and ancient manuscripts strongly support the feasibility of the proposed system. The recognition rates achieve 91.89% (IFN/ENIT) and 94.61% (ancient manuscripts)
Approximate Ensemble Methods for Physical Activity Recognition Applications
Get PDFThe main interest of this thesis is on computational methodologies able to reduce the degree of complexity of learning algorithms and its application to physical activity recognition. Random Projections are used to reduce the learning complexity in Multiple Classier Systems. A new boosting algorithm and a new one-class classication methodology are proposed. In both cases, random projections are used for reducing the dimensionality of the problem and for generating diversity, exploiting in this way the benefits that ensemble learning provides in terms of performances and stability. The practical focus of the thesis is on physical activity recongition using wearable sensors. A new hardware platform for wearable computing application has been developed and used for gathering activity data. Based on the classication methodologies developed and the study conducted on physical activity classication, a machine learning architecture capable to provide a continuous authentication mechanism for mobile-devices users has been designed. The system, based on a personalized classifier, states on the analysis of the characteristic gait patterns typical of each individual ensuring an unobtrusive and continuous authentication mechanism
Automated Classification of Cricket Pitch Frames in Cricket Video
Get PDFAutomated detection of the cricket pitch is a fundamental step in content-based indexing and summarization of cricketvideos. In this paper, we propose visual-content based algorithms to automate the extraction of video frames with thecricket pitch in focus from input cricket videos. As a preprocessing step, we first select a subset of frames with a viewof the cricket field. This reduces the search space by eliminating frames that contain a view of the audience, close-upshots of specific players, advertisements, etc. The subset of frames containing the cricket field is then processed using astatistical modeling of the grayscale (brightness) histogram (SMoG). Since, in the present day, most videos are shot incolor and SMoG does not utilize this information, we propose an alternative: color quantization based region of interestextraction (CQRE). Experimental results demonstrate that successive application of the two methods outperforms eitherone applied exclusively, regardless of the quality of the input. The SMoG-CQRE combination for cricket pitch detectionyields an average accuracy of 98:6% in the best case (a high resolution video with good contrast) and an average accuracyof 87:9% in the worst case (a low resolution video with poor contrast). Since, the extraction of pitch frames only formsthe first step in analyzing key action frames in a match, we also present an an algorithm for player detection in theseframes
Monocular Depth Cues in Computer Vision Applications
Get PDFIn the computer vision field, if image depth information were available, many tasks could be posed from a different perspective for the sake of higher performance and robustness. In our thesis, we have demonstrated how coarse depth information can be integrated in different tasks following alternative strategies to obtain more precise and robust results in three computer vision applications: camera rotation parameters estimation, background estimation and pedestrian candidate generation