No description available.

京都大学 / 0048 / 新制・課程博士 / 博士(医学) / 甲第20259号 / 医博第4218号 / 新制||医||1020(附属図書館) / 京都大学大学院医学研究科医学専攻 / (主査)教授 渡邉 大, 教授 林 康紀, 教授 髙橋 良輔 / 学位規則第4条第1項該当 / Doctor of Medical Science / Kyoto University / DFAM

Visual tracking is one of the most important applications of computer vision. Several tracking systems have been developed which either focus mainly on the tracking of targets moving on a plane, or attempt to reduce the 3-dimensional tracking problem to the tracking of a set of characteristic points of the target. These approaches are seriously handicapped in complex visual situations, particularly those involving significant perspective, textures, repeating patterns, or occlusion. This dissertation describes a new approach to visual tracking for monocular and binocular image sequences, and for both passive and active cameras. The method combines Kalman-type prediction with steepest-descent search for correspondences, using 2-dimensional affine mappings between images. This approach differs significantly from many recent tracking systems, which emphasize the recovery of 3-dimensional motion and/or structure of objects in the scene. We argue that 2-dimensional area-based matching is sufficient in many situations of interest, and we present experimental results with real image sequences to illustrate the efficacy of this approach. Image matching between two images is a simple one to one mapping, if there is no occlusion. In the presence of occlusion wrong matching is inevitable. Few approaches have been developed to address this issue. This dissertation considers the effect of occlusion on tracking a moving object for both monocular and binocular image sequences. The visual tracking system described here attempts to detect occlusion based on the residual error computed by the matching method. If the residual matching error exceeds a user-defined threshold, this means that the tracked object may be occluded by another object. When occlusion is detected, tracking continues with the predicted locations based on Kalman filtering. This serves as a predictor of the target position until it reemerges from the occlusion again. Although the method uses a constant image velocity Kalman filtering, it has been shown to function reasonably well in a non-constant velocity situation. Experimental results show that tracking can be maintained during periods of substantial occlusion. The area-based approach to image matching often involves correlation-based comparisons between images, and this requires the specification of a size for the correlation windows. Accordingly, a new approach based on moment invariants was developed to select window size adaptively. This approach is based on the sudden increasing or decreasing in the first Maitra moment invariant. We applied a robust regression model to smooth the first Maitra moment invariant to make the method robust against noise. This dissertation also considers the effect of spatial quantization on several moment invariants. Of particular interest are the affine moment invariants, which have emerged, in recent years as a useful tool for image reconstruction, image registration, and recognition of deformed objects. Traditional analysis assumes moments and moment invariants for images that are defined in the continuous domain. Quantization of the image plane is necessary, because otherwise the image cannot be processed digitally. Image acquisition by a digital system imposes spatial and intensity quantization that, in turn, introduce errors into moment and invariant computations. This dissertation also derives expressions for quantization-induced error in several important cases. Although it considers spatial quantization only, this represents an important extension of work by other researchers. A mathematical theory for a visual tracking approach of a moving object is presented in this dissertation. This approach can track a moving object in an image sequence where the camera is passive, and when the camera is actively controlled. The algorithm used here is computationally cheap and suitable for real-time implementation. We implemented the proposed method on an active vision system, and carried out experiments of monocular and binocular tracking for various kinds of objects in different environments. These experiments demonstrated that very good performance using real images for fairly complicated situations. / Ph. D.

Recent years have witnessed tremendous success in using deep learning approaches to handle low-level vision problems. Most of the deep learning based methods address the low-level vision problem by training a neural network to approximate the mapping from the inputs to the desired ground truths. However, directly learning this mapping is usually difficult and cannot achieve ideal performance. Besides, under the setting of unsupervised learning, the general deep learning approach cannot be used. In this thesis, we investigate and address several problems in low-level vision using the proposed approaches. To learn a better mapping using the existing data, an indirect domain shift mechanism is proposed to add explicit constraints inside the neural network for single image dehazing. This allows the neural network to be optimized across several identified neighbours, resulting in a better performance. Despite the success of the proposed approaches in learning an improved mapping from the inputs to the targets, three problems of unsupervised learning is also investigated. For unsupervised monocular depth estimation, a teacher-student network is introduced to strategically integrate both supervised and unsupervised learning benefits. The teacher network is formed by learning under the binocular depth estimation setting, and the student network is constructed as the primary network for monocular depth estimation. In observing that the performance of the teacher network is far better than that of the student network, a knowledge distillation approach is proposed to help improve the mapping learned by the student. For single image dehazing, the current network cannot handle different types of haze patterns as it is trained on a particular dataset. The problem is formulated as a multi-domain dehazing problem. To address this issue, a test-time training approach is proposed to leverage a helper network in assisting the dehazing network adapting to a particular domain using self-supervision. In lossy compression system, the target distribution can be different from that of the source and ground truths are not available for reference. Thus, the objective is to transform the source to target under a rate constraint, which generalizes the optimal transport. To address this problem, theoretical analyses on the trade-off between compression rate and minimal achievable distortion are studied under the cases with and without common randomness. A deep learning approach is also developed using our theoretical results for addressing super-resolution and denoising tasks. Extensive experiments and analyses have been conducted to prove the effectiveness of the proposed deep learning based methods in handling the problems in low-level vision. / Thesis / Doctor of Philosophy (PhD)

Certaines recherches ont investigué le traitement visuel de bas et de plus hauts niveaux chez des personnes neurotypiques et chez des personnes ayant un trouble du spectre de l’autisme (TSA). Cependant, l’interaction développementale entre chacun de ces niveaux du traitement visuel n’est toujours pas bien comprise. La présente thèse a donc deux objectifs principaux. Le premier objectif (Étude 1) est d’évaluer l’interaction développementale entre l’analyse visuelle de bas niveaux et de niveaux intermédiaires à travers différentes périodes développementales (âge scolaire, adolescence et âge adulte). Le second objectif (Étude 2) est d’évaluer la relation fonctionnelle entre le traitement visuel de bas niveaux et de niveaux intermédiaires chez des adolescents et des adultes ayant un TSA. Ces deux objectifs ont été évalué en utilisant les mêmes stimuli et procédures. Plus précisément, la sensibilité de formes circulaires complexes (Formes de Fréquences Radiales ou FFR), définies par de la luminance ou par de la texture, a été mesurée avec une procédure à choix forcés à deux alternatives. Les résultats de la première étude ont illustré que l’information locale des FFR sous-jacents aux processus visuels de niveaux intermédiaires, affecte différemment la sensibilité à travers des périodes développementales distinctes. Plus précisément, lorsque le contour est défini par de la luminance, la performance des enfants est plus faible comparativement à celle des adolescents et des adultes pour les FFR sollicitant la perception globale. Lorsque les FFR sont définies par la texture, la sensibilité des enfants est plus faible comparativement à celle des adolescents et des adultes pour les conditions locales et globales. Par conséquent, le type d’information locale, qui définit les éléments locaux de la forme globale, influence la période à laquelle la sensibilité visuelle atteint un niveau développemental similaire à celle identifiée chez les adultes. Il est possible qu’une faible intégration visuelle entre les mécanismes de bas et de niveaux intermédiaires explique la sensibilité réduite des FFR chez les enfants. Ceci peut être attribué à des connexions descendantes et horizontales immatures ainsi qu’au sous-développement de certaines aires cérébrales du système visuel. Les résultats de la deuxième étude ont démontré que la sensibilité visuelle en autisme est influencée par la manipulation de l’information locale. Plus précisément, en présence de luminance, la sensibilité est seulement affectée pour les conditions sollicitant un traitement local chez les personnes avec un TSA. Cependant, en présence de texture, la sensibilité est réduite pour le traitement visuel global et local. Ces résultats suggèrent que la perception de formes en autisme est reliée à l’efficacité à laquelle les éléments locaux (luminance versus texture) sont traités. Les connexions latérales et ascendantes / descendantes des aires visuelles primaires sont possiblement tributaires d’un déséquilibre entre les signaux excitateurs et inhibiteurs, influençant ainsi l’efficacité à laquelle l’information visuelle de luminance et de texture est traitée en autisme. Ces résultats supportent l’hypothèse selon laquelle les altérations de la perception visuelle de bas niveaux (local) sont à l’origine des atypies de plus hauts niveaux chez les personnes avec un TSA. / Most studies investigating visual perception in typically developing populations and in Autism Spectrum Disorder (ASD) have assessed lower- (local) and higher-levels (global) of processing in isolation. However, much less is known about the developmental interactions between mechanisms mediating early- and intermediate-level vision in both typically developing populations and in ASD. Based on such premise, the present thesis had two main objectives. The first objective (Study 1) was to evaluate the developmental interplay between low- and intermediate-levels of visual analysis at different periods of typical development (school-age, adolescence and adulthood). The second objective (Study 2) was to evaluate the functional relationship between low- and intermediate-levels of visual analysis in adolescents and adults diagnosed with ASD. Common methodologies were used to assess both objectives. Specifically, sensitivity to slightly curved circles (Radial Frequency Patterns or RFP), defined by luminance or texture information, was measured using a two alternative temporal forced choice procedure. Results obtained in Study 1 demonstrated that local information defining a RFP (mediated by intermediate visual mechanisms) differentially affected sensitivity at different periods of development. Specifically, when the contour was luminance-defined, children performed worse when compared to adolescents and adults only when RFPs targeted a global processing style (few deformations along the RFP’s contour). When RFPs were texture-defined, children’s sensitivity was worse compared to that of adolescents and adults for both local and global conditions. Therefore, timing of adult-like sensitivity to RFPs is dependent on the type of local physical elements defining its global shape. Poor visual integration between low and intermediate visual mechanisms, which could be attributed to immature feedback and horizontal connections as well as under-developed visual cortical areas, may account for such reduced sensitivity in children. Results obtained from Study 2 demonstrated that manipulating the local physical elements of RFPs impacts visual sensitivity in ASD. Specifically, sensitivity to RFPs is unaffected in ASD only when visual analysis is dependent on local deformations of luminance-defined contours. However, sensitivity is reduced for both local and global visual analysis when shapes are texture-defined. Such results suggest that intermediate-level, shape perception in ASD is functionally related to the efficacy with which local physical elements (luminance versus texture) are processed. It is possible that abnormal lateral or feed-forward / feedback connectivity within primary visual areas in ASD, which possibly arise from excitatory / inhibitory signalling imbalance, accounts for differential efficacy with which luminance and texture information is processed in ASD. These results support the hypothesis that atypical higher-level perception in ASD, when present, may have early (local) visual origins.