Abstract: More realistic and accurate models of the scene illumination and object reflectance can greatly improve the quality of many computer vision and computer graphics tasks. Using such model, a more profound knowledge about the interaction of light with object surfaces can be established which proves crucial to a variety of computer vision applications. In the current work, we investigate the various existing approaches to illumination and reflectance modeling and form an analysis on their shortcomings in capturing the complexity of real-world scenes. Based on this analysis we propose improvements to different aspects of reflectance and illumination estimation in order to more realistically model the real-world scenes in the presence of complex lighting phenomena (i.e, multiple illuminants, interreflections and shadows). Moreover, we captured our own multi-illuminant dataset which consists of complex scenes and illumination conditions both outdoor and in laboratory conditions. In addition we investigate the use of synthetic data to facilitate the construction of datasets and improve the process of obtaining ground-truth information.

Abstract: Motion perception is one of the most important attributes of the human brain. Visual motion perception consists in inferring speed and direction of elements in a scene based on visual inputs. Analogously, computer vision is assisted by motion cues in the scene. Motion detection in computer vision is useful in solving problems such as segmentation, depth from motion, structure from motion, compression, navigation and many others. These problems are common in several applications, for instance, video surveillance, robot navigation and advanced driver assistance systems (ADAS). One of the most widely used techniques for motion detection is the optical flow estimation. The work in this thesis attempts to make optical flow suitable for the requirements and conditions of driving scenarios. In this context, a novel space-variant representation called reverse log-polar representation is proposed that is shown to be better than the traditional log-polar space-variant representation for ADAS. The space-variant representations reduce the amount of data to be processed. Another major contribution in this research is related to the analysis of the influence of specific characteristics from driving scenarios on the optical flow accuracy. Characteristics such as vehicle speed and
road texture are considered in the aforementioned analysis. From this study, it is inferred that the regularization weight has to be adapted according to the required error measure and for different speeds and road textures. It is also shown that polar represented optical flow suits driving scenarios where predominant motion is translation. Due to the requirements of such a study and by the lack of needed datasets a new synthetic dataset is presented; it contains: i) sequences of different speeds and road textures in an urban scenario; ii) sequences with complex motion of an on-board camera; and iii) sequences with additional moving vehicles in the scene. The ground-truth optical flow is generated by the ray-tracing technique. Further, few applications of optical flow in ADAS are shown. Firstly, a robust RANSAC based technique to estimate horizon line is proposed. Then, an egomotion estimation is presented to compare the proposed space-variant representation with the classical one. As a final contribution, a modification in the regularization term is proposed that notably improves the results
in the ADAS applications. This adaptation is evaluated using a state of the art optical flow technique. The experiments on a public dataset (KITTI) validate the advantages of using the proposed modification.

Abstract: Natural languages can often be modelled by suitable grammars whose knowledge can improve the word spotting results. The implicit contextual information is even more useful when dealing with information that is intrinsically described as one collection of records. In this paper, we present one approach to word spotting which uses the contextual information of records to improve the results. The method relies on Markov Logic Networks to probabilistically model the relational organization of handwritten records. The performance has been evaluated on the Barcelona Marriages Dataset that contains structured handwritten records that summarize marriage information.

Abstract: Obstacle avoidance remains a challenging task for Micro Aerial Vehicles (MAV), due to their limited payload capacity to carry advanced sensors. Unlike larger vehicles, MAV can only carry light weight sensors, for instance a camera, which is our main assumption in this work. We explore passive monocular depth estimation and propose a novel method Position Aided Depth Estimation
(PADE). We analyse PADE performance and compare it against the extensively used Time To Collision (TTC). We evaluate the accuracy, robustness to noise and speed of three Optical Flow (OF) techniques, combined with both depth estimation methods. Our results show PADE is more accurate than TTC at depths between 0-12 meters and is less sensitive to noise. Our findings highlight the potential application of PADE for MAV to perform safe autonomous navigation in
unknown and unstructured environments.

Abstract: Otolith end organs of vertebrates sense linear accelerations of the head and gravitation. The hair cells on their epithelia are responsible for transduction. In mammals, the striola, parallel to the line where hair cells reverse their polarization, is a narrow region centered on a curve with curvature and torsion. It has been shown that the striolar region is functionally different from the rest, being involved in a phasic vestibular pathway. We propose a mathematical and computational model that explains the necessity of this amazing geometry for the striola to be able to carry out its function. Our hypothesis, related to the biophysics of the hair cells and to the physiology of their afferent neurons, is that striolar afferents collect information from several type I hair cells to detect the jerk in a large domain of acceleration directions. This predicts a mean number of two calyces for afferent neurons, as measured in rodents. The domain of acceleration directions sensed by our striolar model is compatible with the experimental results obtained on monkeys considering all afferents. Therefore, the main result of our study is that phasic and tonic vestibular afferents cover the same geometrical fields, but at different dynamical and frequency domains.

Abstract: Pedestrian detection is of paramount interest for many applications, e.g. Advanced Driver Assistance Systems, Intelligent Video Surveillance and Multimedia systems. Most promising pedestrian detectors rely on appearance-based classifiers trained with annotated data. However, the required annotation step represents an intensive and subjective task for humans, what makes worth to minimize their intervention in this process by using computational tools like realistic virtual worlds. The reason to use these kind of tools relies in the fact that they allow the automatic generation of precise and rich annotations of visual information. Nevertheless, the use of this kind of data comes with the following question: can a pedestrian appearance model learnt with virtual-world data work successfully for pedestrian detection in real-world scenarios?. To answer this question, we conduct different experiments that suggest a positive answer. However, the pedestrian classifiers trained with virtual-world data can suffer the so called dataset shift problem as real-world based classifiers does. Accordingly, we have designed different domain adaptation techniques to face this problem, all of them integrated in a same framework (V-AYLA). We have explored different methods to train a domain adapted pedestrian classifiers by collecting a few pedestrian samples from the target domain (real world) and combining them with many samples of the source domain (virtual world). The extensive experiments we present show that pedestrian detectors developed within the V-AYLA framework do achieve domain adaptation. Ideally, we would like to adapt our system without any human intervention. Therefore, as a first proof of concept we also propose an unsupervised domain adaptation technique that avoids human intervention during the adaptation process. To the best of our knowledge, this Thesis work is the first demonstrating adaptation of virtual and real worlds for developing an object detector. Last but not least, we also assessed a different strategy to avoid the dataset shift that consists in collecting real-world samples and retrain with them in such a way that no bounding boxes of real-world pedestrians have to be provided. We show that the generated classifier is competitive with respect to the counterpart trained with samples collected by manually annotating pedestrian bounding boxes. The results presented on this Thesis not only end with a proposal for adapting a virtual-world pedestrian detector to the real world, but also it goes further by pointing out a new methodology that would allow the system to adapt to different situations, which we hope will provide the foundations for future research in this unexplored area.

Abstract: Person re-identification is about recognizing people who have passed by a sensor earlier. Previous work is mainly based on RGB data, but in this work we for the first time present a system where we combine RGB, depth, and thermal data for re-identification purposes. First, from each of the three modalities, we obtain some particular features: from RGB data, we model color information from different regions of the body, from depth data, we compute different soft body biometrics, and from thermal data, we extract local structural information. Then, the three information types are combined in a joined classifier. The tri-modal system is evaluated on a new RGB-D-T dataset, showing successful results in re-identification scenarios.