Electronic Letters on Computer Vision and Image Analysis (ELCVIA - Universitat Autònoma de Barcelona)
Not a member yet
343 research outputs found
Sort by
An Adaptive and Integrated Multimodal Sensing And Processing Framework For Long-Range Moving Object Detection And Classification
Full text linkIn applications such as surveillance, inspection and traffic monitoring, long-range detection and classification of targets (vehicles, humans, etc) is a highly desired feature for a sensing system. A single modality will no longer provide the required performance due to the challenges in detection and classification with low resolutions, noisy sensor signals, and various environmental factors due to large sensing distances. Multimodal sensing and processing, on the other hand, can provide complementary information from heterogeneous sensor modalities, such as audio, visual and range sensors. However, there is a lack of effective sensing mechanisms and systematic approaches for sensing and processing using multimodalities. In this thesis, a systematical framework is proposed for Adaptive and Integrated Multimodal Sensing and Processing (AIM-SP) that integrates novel multimodal long-range sensors, adaptive feature selection and learning-based object detection and classification for achieving the goal of adaptive and integrated multimodal sensing and processing. Based on the AIM-SP framework, we have made three unique contributions. First, we have designed a novel multimodal sensor system called Vision-Aided Automated Vibrometry (VAAV), consists of a laser Doppler vibrometer (LDV) and a pair of pan-tilt-zoom (PTZ) cameras, and the system is capable of automatically obtaining visual, range and acoustic signatures for moving object detection at a large distance. It provides a close loop adaptive sensing that allows determination of good surface points and quickly focusing the laser beam of the LDV based on the target detection, surface selection, and distance measurements by the PTZ pair and acoustic signal feedbacks of the LDV. Second, multimodal data of vehicles on both local roads and highways, acquired from multiple sensing sources, are integrated and represented in a Multimodal Temporal Panorama (MTP) for easy alignment and fast labelling of the multimodal data: visual, audio and range. Accuracy of target detection can be improved using multimodalities, and a visual reconstruction method is developed to remove occlusions, motion blurs and perspective distortions of moving vehicles so that scale- and perspective-invariant visual vehicle features are obtained. The concept of MTP is not limited to visual and audio information, but is also applicable when other modalities are available that can be presented in the same time axis. With various types of features extracted on aligned multimodal samples, we made our third contribution on feature modality selection using two approaches. The first approach uses multi-branch sequential-based feature searching (MBSF) and the second one uses boosting-based feature learning (BBFL). In our implementations, three types of visual features are used: aspect ratio and size (ARS), histograms of oriented gradients (HOGs), shape profile (SP), representing simple global scale features, statistical features, and global structure features, respectively. The audio features include short time energy (STE), spectral features (SPECs) which consists of spectral energy, entropy, flux and centroid, and perceptual features (PERCs) are Mel-frequency cepstral coefficients (MFFCs) for the perceptual features. The effectiveness of multimodal feature selection is thoroughly studied through empirical studies. The performance between MBSF and BBFL is compared based on our own dataset, which contains over 3000 samples of mainly four types of moving vehicles: sedans, pickup-trucks, vans and buses under various conditions. From this dataset, a subset of 667 samples of multimodal vehicle data is made publicly available at: http://www.cse.ohio-stata.edu/otcbvs-bench/. A number of important observations on the strengths and weakness of those features and their combinations are made as well
Enhanced Rotational Feature Points Matching using Orientation Correction
Full text linkSeveral techniques have been developed for estimating orientation assignment to make feature points invariant to the rotation for the purpose of matching. However, imperfect estimation of the orientation assignment may lead to feature mismatching and low number of correctly matched points. Besides, several possible candidates with high correlation values for one feature in the reference image may lead to matching confusion. In this paper, we propose a post-processing matching technique to increase the number of correctly matched points and at the same time solve these two issues. The key idea is to correct the orientation of features based on the relative rotational degree between two images which is estimated based on the orientation difference between major correctly matched points after first round matching. Our analysis of the proposed method shows that the number of correctly matched points can be increased up to 50% of the detected points in the reference image. In addition, some mismatched points due to similar correlation value in first round matching can be corrected. Moreover, the proposed algorithm can be applied to different states-of-the-art orientation assignment techniques
Overcomplete Image Representations for Texture Analysis
Full text linkIn recent years, computer vision has played an important role in many scientific and technological areas mainlybecause modern society highlights vision over other senses. At the same time, application requirements and complexity have also increased so that in many cases the optimal solution depends on the intrinsic charac-teristics of the problem; therefore, it is difficult to propose a universal image model. In parallel, advances in understanding the human visual system have allowed to propose sophisticated models that incorporate simple phenomena which occur in early stages of the visual system. This dissertation aims to investigate characteristicsof vision such as over-representation and orientation of receptive fields in order to propose bio-inspired image models for texture analysi
Monitoring and Diagnosing Neonatal Seizures by Video Signal Processing
Full text linkIn this thesis we consider the use of well-known statistical methods to early diagnose, through wire-free low-cost video processing, the potential presence of seizures. For this purpose several approaches, have been proposed: periodicity-based, classification-based and clustering-based approaches
Multi-class learning for vessel characterisation in intravascular ultrasound
Full text linkIn this thesis we tackle the problem of automatic characterization of human coronary vessel in IntravascularUltrasound (IVUS) image modality. The basis for the whole characterization process is machinelearning applied to multi-class problems. In all the presented approaches, the Error-Correcting Output Codes(ECOC) framework is used as central element for the design of multi-class classifiers. Two main contributionsare presented in this thesis. First, a novel method for the design of potential function for DiscriminativeRandom Fields, namely ECOC-DRF, is presented. The method is successfully applied to problems of objectclassification and segmentation in synthetic and natural images. Furthermore, ECOC-DRF is applied toobtain a robust vessel characterization in IVUS image sequences. Based on ECOC-DRF, the main regionsof the coronary artery are robustly segmented by means of a novel holistic approach, namely HoliMAb, representingthe second contribution of this thesis. The HoliMAb framework is applied to problems of lumenborder and media-adventitia border detection, achieving an error comparable with inter-observer variabilityand with state of the art methods
Methods for text segmentation from scene images
Full text linkCamera-captured scene/born-digital image analysis helps in the development of vision for robots to read text, transliterate or translate text, navigate and retrieve search results. However, text in such images does nor follow any standard layout, and its location within the image is random in nature. In addition, motion blur, non-uniform illumination, skew, occlusion and scale-based degradations increase the complexity in locating and recognizing the text in a scene/born-digital image. OTCYMIST method is proposed to segment text from the born-digital images. This method won the first place in ICDAR 2011 and placed in the third position in ICDAR 2013 for its performance on the text segmentation task in robust reading competitions for born-digital image data set. Here, Otsu’s binarization and Canny edge detection are separately carried out on the three colour planes of the image. Connected components (CC’s) obtained from the segmented image are pruned based on thresholds applied on their area and aspect ratio. CC’s with sufficient edge pixels are retained. The centroids of the individual CC’s are used as nodes of a graph. A minimum spanning tree is built using these nodes of the graph. Long edges are broken from the minimum spanning tree of the graph. Pairwise height ratio is used to remove likely non-text components. CC’s are grouped based on their proximity in the horizontal direction to generate bounding boxes (BB’s) of text strings. Overlapping BB’s are removed using an overlap area threshold. Non-overlapping and minimally overlapping BB’s are retained for text segmentation. These BB’s are split vertically to localize text at the word level. A word cropped from a document image can easily be recognized using a traditional optical character recognition (OCR) engine. However, recognizing a word, obtained by manually cropping a scene/born-digital image, is not trivial. Existing OCR engines do not handle these kinds of scene word images effectively. Our intention is to first segment the word image and then pass it to the existing OCR engines for recognition. It is advantageous in two aspects: it avoids building a character classifier from scratch and reduces the word recognition task to a word segmentation task. Here, we propose three bottom-up approaches to segment a cropped word image. These approaches choose different features at the initial stage of segmentation. Power-law transform (PLT) was applied to the pixels of the gray scale born-digital images to non-linearly enhance the histogram. The recognition rate achieved on born-digital word images is 82.9%, which is 20% more than the top performing entry (61.5%) in ICDAR 2011 robust reading competition. The recognition rate is 82.7% and 64.6% for born-digital and scene images of ICDAR 2013 robust reading competition, respectively, using PLT.In addition, we applied PLT to the colour planes such as red, green, blue, intensity and lightness plane by varying the gamma value. We call this technique as Nonlinear enhancement and selection of plane (NESP) for optimal segmentation, which is an improvement over PLT. NESP chooses a particular plane with a proper gamma value based on Fisher discrimination factor. The recognition rate is 72.8% for scene images of ICDAR 2011 robust reading competition, which is 30% higher than the best entry (41.2%). The recognition rate is 81.7% and 65.9% for born-digital and scene images of ICDAR 2013 robust reading competition, respectively, using NESP.Another technique, midline analysis and propagation of segmentation (MAPS), has also been proposed for word segmentation. Here, the middle row pixels of the gray scale image are first segmented and the statistics of the segmented pixels are used to assign text and non-text labels to the rest of the image pixels using min-cut method. Gaussian model is fitted on the middle row segmented pixels before the assignment of other pixels. In MAPS method, we assume the middle row pixels are least affected by any of the degradations. This assumption is validated by the good word recognition rate of 71.7% on ICDAR 2011 robust reading competition for scene images. The recognition rate is 83.8% and 66.0% for born-digital and scene images of ICDAR 2013 robust reading competition, respectively, using MAPS. The best reported results for ICDAR 2003 word images is 61.1% using custom lexicons containing the list of test words. On the other hand, NESP and MAPS achieve 66.2% and 64.5% for ICDAR 2003 word images without using any lexicon. By using similar custom lexicon, the recognition rates for ICDAR 2003 word images go up to 74.9% and 74.2% for NESP and MAPS methods, respectively.We manually segmented word images and recognized these images using OCR to benchmark maximum possible recognition rate for each database. The recognition rates of the proposed methods and the benchmark results are reported on the seven publicly available word image data sets and compared with the results reported in the literature.We have designed a classifier to recognize Kannada characters and words from Chars74k data set and our own image collection, respectively. Discrete cosine transform (DCT) and block DCT are used as features to train separate classifiers. Kannada words are segmented using the same techniques (MAPS and NESP) and further segmented into groups of components, since a Kannada character may be represented by a single component or a group of components in an image. The recognition rate on Kannada words is reported for different features with and without the use of a lexicon. The obtained recognition performance for Kannada character recognition (11.4%) is three times the best performance (3.5%) reported in the literature.This thesis has dealt with the principal aspects of camera captured scene/born-digital text image analysis: text localization, text segmentation, and word recognition. We have benchmarked the recognition rates of five word image data sets. We conducted a multi-script robust reading competition as part of ICDAR 2013. This competition was aimed to determine whether the text localization and segmentation methods were capable of handling any text, independent of the script
A Parallel Framework for Video Super-resolution
Full text linkIn this work we propose a framework for increasing the processing efficiency of super-resolution algorithms. The framework is targeted at super-resolution video processing algorithms, that require a large amount of data processing. We propose a set of strategies that use a combination of data simplification and parallel processing. The simplification strategies are used to decrease the amount of complex data and, consequently, decrease the processing time. The parallel processing strategies are designed so that major modifications of the super-resolution algorithms are not required. As presented in this work, the framework is fast and makes the video resolution increase timely
Automatic Segmentation of Anatomical Structures using Deformable Models and Bio-Inspired/Soft Computing
Full text linkThis PhD dissertation is focused on the development of algorithms for the automatic segmentation of anatomical structures in biomedical images, usually the hippocampus in histological images from the mouse brain. Such algorithms are based on computer vision techniques and artificial intelligence methods. More precisely, on the one hand, we take advantage of deformable models to segment the anatomical structure under consideration, using prior knowledge from different sources, and to embed the segmentation into an optimization framework. On the other hand, metaheuristics and classifiers can be used to perform the optimization of the target function defined by the shape model (as well as to automatically tune the system parameters), and to refine the results obtained by the segmentation process, respectively. Three new different methods, with their corresponding advantages and disadvantages, are described and tested. A broad theoretical discussion, together with an extensive introduction to the state of the art, has also been included to provide an overview necessary for understanding the developed methods
Moving Cast Shadow Detection
No full textMotion perception is an amazing innate ability of the creatures on the planet. This adroitness entails a functional advantage that enables species to compete better in the wild. The motion perception ability is usually employed at different levels, allowing from the simplest interaction with the ’physis’ up to the most transcendental survival tasks. Among the five classical perception system , vision is the most widely used in the motion perception field. Millions years of evolution have led to a highly specialized visual system in humans, which is characterized by a tremendous accuracy as well as an extraordinary robustness. Although humans and an immense diversity of species can distinguish moving object with a seeming simplicity, it has proven to be a difficult and non trivial problem from a computational perspective.In the field of Computer Vision, the detection of moving objects is a challenging and fundamental research area. This can be referred to as the ’origin’ of vast and numerous vision-based research sub-areas. Nevertheless, from the bottom to the top of this hierarchical analysis, the foundations still relies on when and where motion has occurred in an image.Pixels corresponding to moving objects in image sequences can be identified by measuring changes in their values. However, a pixel’s value (representing a combination of color and brightness) could also vary due to other factors such as: variation in scene illumination, camera noise and nonlinear sensor responses among others. The challenge lies in detecting if the changes in pixels’ value are caused by a genuine object movement or not. An additional challenging aspect in motion detection is represented by moving cast shadows. The paradox arises because a moving object and its cast shadow share similar motion patterns. However, a moving cast shadow is not a moving object. In fact, a shadow represents a photometric illumination effect caused by the relative position of the object with respect to the light sources.Shadow detection methods are mainly divided in two domains depending on the application field. One normally consists of static images where shadows are casted by static objects, whereas the second one is referred to image sequences where shadows are casted by moving objects. For the first case, shadows can provide additional geometric and semantic cues about shape and position of its casting object as well as the localization of the light source. Although the previous information can be extracted from static images as well as video sequences, the main focus in the second area is usually change detection, scene matching or surveillance. In this context, a shadow can severely affect with the analysis and interpretation of the scene.The work done in the thesis is focused on the second case, thus it addresses the problem of detection and removal of moving cast shadows in video sequences in order to enhance the detection of moving object
Adaptive Texture Description and Estimation of the Class Prior Probabilities for Seminal Quality Control
Full text linkMotivation of the Thesis Semen quality assessment is a crucial task in artificial insemination (AI) processes, both human and animal. Animal AI allows farmers to save time and money (e.g. working with a limited number of animals). They purchase semen samples to companies, which have to carry out strict quality controls to guarantee that they are optimal for fertilization. A sample with a high proportion of (i) dead spermatozoa, or (ii) sperm heads with damaged acrosomes (the acrosome is a membrane that covers the anterior part of the sperm head and makes possible the penetration into the ovum) will have low fertilization potential. Therefore, sperm vitality and acrosome integrity are two of the parameters assessed by veterinaries in semen quality control processes. Both are assessed by means of a visual process which entails expensive equipments (stains and fluorescence microscopes) and may be a source of errors, as any manual process is. The contributions in the field of Image Processing and Machine Learning made on this PhD. Thesis [2] may be used to develop an automatic process to assess the proportions of damaged acrosomes or dead spermatozoa using just a phase contrast microscope (which almost any lab has) and a digital camera. Concretely, several texture description approaches have been evaluated. In addition, a new intelligent segmentation process, an adaptive texture description method, and two robust approaches for estimating class proportions of unlabelled datasets have been proposed. All these methods are applied to automatic boar semen quality estimation