This work deals with the OpenCL technology and its use for the task of object detection. The introduction is devoted to description of OpenCL fundamentals, as well as basic theory of object detection. Next chapter of the work is analysis, with design proposal which takes into consideration the possibilities of OpenCL. Further, there's description of implementation of detection application and experimental evaluation of detector's performance. The last chapter summarizes the achieved results.

This master's thesis deals with face detection on mobile phones with Android OS. The introduction describes some algorithms used for pattern detection from image, as well as various techniques of features extracting. After that Android platform development specifics, including basic description of development tools, are described. Architecture of SIMD is introduced in next part of this work. After acquiring basic knowleage analysis and implementation of final app are descrited. Performance tests are conducted whose results are summarized in the conclusion.

This thesis deals with moving objects detection in video sequences. The principal aim of such detection is to detect and locate motion in the image, separate individual objects, and track these objects. Subsequently, to eliminate shadows, the paper introduces method of motion detection based on Local Binary Patterns together with differential method above the HSV color space. The proposed method provides rapid and accurate movement detection in video sequences.

Aim of this project is to evaluate effectivity of various texture features within the context of image processing, particulary the task of texture recognition and classification. My work focuses on comparing and discussion of usage and efficiency of texture features based on local binary patterns and co- ccurence matrices. As classification algorithm is concerned, cluster analysis was choosen.

Ces technologies ont poussé les chercheurs à imaginer la possibilité d'automatiser et émuler les capacités de perception visuels des animaux et de l'homme lui-même. Depuis quelques décennies le domaine de la vision par ordinateur a essayé plusieurs approches et une vaste gamma d'applications a été développée avec un succès partielle: la recherche des images basé sur leur contenu, la exploration de donnés à partir des séquences vidéo, la ré-identification des objets par des robots, etc. Quelques applications sont déjà sur le marché et jouissent déjà d'un certain succès commercial.La reconnaissance visuelle c'est un problème étroitement lié à l'apprentissage de catégories visuelles à partir d'un ensemble limité d'instances. Typiquement deux approches sont utilisées pour résoudre ce problème: l'apprentissage des catégories génériques et la ré-identification d'instances d'un objet un particulière. Dans le dernier cas il s'agit de reconnaître un objet ou personne en particulière. D'autre part, la reconnaissance générique s'agit de retrouver tous les instances d'objets qui appartiennent à la même catégorie conceptuel: tous les voitures, les piétons, oiseaux, etc.Cette thèse propose un système de vision par ordinateur capable de détecter et suivre plusieurs objets dans les séquences vidéo. L'algorithme pour la recherche de correspondances proposé se base sur les matrices de covariance obtenues à partir d'un ensemble de propriétés des images (couleur et texture principalement). Son principal avantage c'est qu'il utilise un descripteur qui permet l'introduction des sources d'information très hétérogènes pour représenter les cibles. Cette représentation est efficace pour le suivi d'objets et son ré-identification.Quatre contributions sont introduites dans cette thèse.Tout d'abord cette thèse s'intéresse à l'invariance des algorithmes de suivi face aux changements du contexte. Nous proposons ici une méthodologie pour mesurer l’importance de l'information couleur en fonction de ses niveaux d’illumination et saturation. Puis, une deuxième partie se consacre à l'étude de différentes méthodes de suivi, ses avantages et limitations en fonction du type d'objet à suivre (rigide ou non rigide par exemple) et du contexte (caméra statique ou mobile). Le méthode que nous proposons s'adapte automatiquement et utilise un mécanisme de commutation entre différents méthodes de suivi qui considère ses qualités complémentaires. Notre algorithme se base sur un modèle de covariance qui fusionne les informations couleur-texture et le flot optique (KLT) modifié pour le rendre plus robuste et adaptable face aux changements d’illumination. Une deuxième approche se appuie sur l'analyse des différents espaces et invariants couleur à fin d'obtenir un descripteur qui garde un bon équilibre entre pouvoir discriminant et robustesse face aux changements d'illumination.Une troisième contribution porte sur le problème de suivi multi-cibles ou plusieurs difficultés apparaissent : la confusion d'identités, les occultations, la fusion et division des trajectoires-détections, etc.La dernière partie se consacre à la vitesse des algorithmes à fin de fournir une solution rapide et utilisable dans les applications embarquées. Cette thèse propose une série d'optimisations pour accélérer la mise en correspondance à l'aide de matrices de covariance. Transformations de mise en page de données, la vectorisation des calculs (à l'aide d'instructions SIMD) et certaines transformations de boucle permettent l'exécution en temps réel de l'algorithme non seulement sur les grands processeurs classiques de Intel, mais aussi sur les plateformes embarquées (ARM Cortex A9 et Intel U9300). / Visual recognition is the problem of learning visual categories from a limited set of samples and identifying new instances of those categories, the problem is often separated into two types: the specific case and the generic category case. In the specific case the objective is to identify instances of a particular object, place or person. Whereas in the generic category case we seek to recognize different instances that belong to the same conceptual class: cars, pedestrians, road signs and mugs. Specific object recognition works by matching and geometric verification. In contrast, generic object categorization often includes a statistical model of their appearance and/or shape.This thesis proposes a computer vision system for detecting and tracking multiple targets in videos. A preliminary work of this thesis consists on the adaptation of color according to lighting variations and relevance of the color. Then, literature shows a wide variety of tracking methods, which have both advantages and limitations, depending on the object to track and the context. Here, a deterministic method is developed to automatically adapt the tracking method to the context through the cooperation of two complementary techniques. A first proposition combines covariance matching for modeling characteristics texture-color information with optical flow (KLT) of a set of points uniformly distributed on the object . A second technique associates covariance and Mean-Shift. In both cases, the cooperation allows a good robustness of the tracking whatever the nature of the target, while reducing the global execution times .The second contribution is the definition of descriptors both discriminative and compact to be included in the target representation. To improve the ability of visual recognition of descriptors two approaches are proposed. The first is an adaptation operators (LBP to Local Binary Patterns ) for inclusion in the covariance matrices . This method is called ELBCM for Enhanced Local Binary Covariance Matrices . The second approach is based on the analysis of different spaces and color invariants to obtain a descriptor which is discriminating and robust to illumination changes.The third contribution addresses the problem of multi-target tracking, the difficulties of which are the matching ambiguities, the occlusions, the merging and division of trajectories.Finally to speed algorithms and provide a usable quick solution in embedded applications this thesis proposes a series of optimizations to accelerate the matching using covariance matrices. Data layout transformations, vectorizing the calculations (using SIMD instructions) and some loop transformations had made possible the real-time execution of the algorithm not only on Intel classic but also on embedded platforms (ARM Cortex A9 and Intel U9300).

近紅外線影像由於其物理特性與成像方式，其紋理細節都有發散模糊不清的現象，對於以紋理為主要特徵的年齡辨識問題而言更具挑戰。本論文主要目的是以近紅外線人臉影像為基礎，找出對近紅外線年齡特徵有最佳描述力的特徵描述子，辨識近紅外線影像中被拍攝者的年齡區間，建構整個年齡層估算機制。 相關研究一部分關注可見光年齡辨識，另一部分則聚焦在近紅外線人臉辨識，目前還沒有近紅外線年齡辨識的相關文獻能參考，如何從接近的研究領域找尋是適當的演算法是本研究遇到的第一個挑戰。在資料庫的部分，FGNET和MORPH常被用於可見光的年齡辨識議題； PolyU和LDHF則用於近紅外線人臉辨識相關研究，在目前沒有近紅外線年齡資料庫的情況，本研究自建RSNIR（Intel RealSense Near-Infrared Age Database），因此如何標準化拍攝環境流程、蒐集穩定的近紅外線影像是本研究面臨的第二個挑戰。 區域性特徵擷取方法的關鍵在於特徵描述子的描述力。本研究以LBP（Local Binary Patterns）為基礎，探討LBP在內的24個特徵描述子，最後實驗測試各個描述子在RSNIR的辨識率，結果發現基本型Fuzzy LBP和擴充型RILBP對近紅外線年齡特徵有最佳描述力。在空間譜子區塊（patch）設計部分，以3x3切割子區塊數的辨識效果最好，反應出其與影像校正時的人臉影像空間定義方式有關。

<p>Problemi klasifikacije i segmentacije digitalnih slika su veoma<br />aktuelni i zastupljeni u praksi. Potreba za modelima koji razmatraju<br />ovu problematiku u poslednjih nekoliko decenija ubrzanim tempom<br />poprima sve veći značaj i obim u svakodnevnom životu. Koriste se u<br />računarskoj grafici, prepoznavanju oblika, medicinskoj analizi slika,<br />saobraćaju, analizi dokumenata, pokreta i izraza lica i sl.<br />U okviru ove disertacije, predstavljeno istraživanje motivisano je<br />primenama razvijenih modela u klasifikaciji i segmentaciji<br />digitalnih slika. Istraživanje obuhvata dva segmenta. Ovi segmenti<br />povezani su terminom neodređenosti, koji je uz upotrebu adekvatnog<br />matematičkog aparata (teorije fazi skupova), ugrađen u modele razvije<br />za primenu u obradi slike.<br />Jedan pravac istraživanja baziran je na teoriji fazi skupova, t-<br />normama, t-konormama, operatorima agregacije i agregiranim<br />funkcijama rastojanja. U okviru toga, istraživanje je sprovedeno sa<br />struktuiranom matematičkom podlogom, izložene su osnovne<br />definicije, teoreme, kao i osobine korištenih operatora, prošireni<br />su teorijski koncepti t-normi i t-konormi. Definisani su novi tipovi<br />operatora agregacije i njihovom primenom konstruisane su nove<br />funkcije rastojanja, čija je upotreba diskutovana kroz uspešnost u<br />procesu segmentacije digitalnih slika.<br />Drugi pravac istraživanja, izložen u ovoj disertaciji, obuhvata više<br />inženjerski pristup rešavanju problema klasifikacije tekstura<br />digitalnih slika. U skladu sa tim, detaljno je analizirana i<br />diskutovana klasa lokalnih binarnih deskriptora teksture.<br />Inspirisana uspešnošću pomenute LBP klase deskriptora, uvedena je<br />jedna nova podfamilija &alpha;-deskriptora teksture. Uvedeni model<br />deskriptora formiran je na temeljima idejnih principa lokalnih<br />binarnih kodova i bazičnih pojmova iz teorije fazi skupova. Praktična<br />upotreba i značaj predstavljenog modela demonstrirani su kroz veoma<br />uspešne procese klasifikacije na nekoliko javno dostupnih baza slika.</p> / <p>Classification and segmentation problems of digital images is a very attractive<br />topic and has been making impact in many different applied disciplines. In the<br />past few decades, the demand for models that address these issues has been<br />gaining momentum and applications in everyday life. These models are used in<br />computer graphics, shape recognition, medical image analysis, traffic, document<br />analysis, facial movements and expressions, etc.<br />The research within this doctoral dissertation was motivated by the application of<br />developed methods in classification and segmentation tasks. The conducted<br />research covered two segments, which were linked by the term of indeterminacy,<br />with the usage of the theory of fuzzy sets, which is incorporated into methods<br />developed for application in image processing.<br />One direction of the research was founded on the theory of fuzzy sets, t-norms,<br />t-conorms, aggregation operators, and aggregated distance functions. Within this<br />framework, the research was conducted with a structured mathematical<br />background. Firstly, basic definitions, theorems and characteristics of the used<br />operators were presented, followed by the theoretical concepts of t-norms and tconorms<br />that were extended. New types of aggregation operators and distance<br />functions were defined, and finally, their contribution in the digital image<br />segmentation process was explored and discussed.<br />The second direction of the research presented in this dissertation involved more<br />of an engineering-type of approach to solving the problem of the classification of<br />digital image textures. To that end, a class of local binary texture descriptors<br />(LBPs) was analyzed and discussed in detail. Inspired by the results of the<br />above-mentioned LBP descriptors, one new sub-family of the $\alpha$-<br />descriptors was introduced by the author. The introduced descriptor model was<br />based on the conceptual principles of LBPs and basic definitions from the fuzzy<br />set theory. Its practical usage and importance were established and reflected in<br />very successful classification results, achieved in the application on several<br />publicly available image datasets.</p>

This master’s Thesis focuses on image processing on Android platform and development of an application, that is able to do face detection and recognition in real scene. Thesis gives highlight of modern algorithms of face detection. It first examines and compares the standard features of Android platform (FaceDetector a FaceDetectionListener) and JJIL, OpenIMAJ, OpenCV libraries experiment, and presents the results. For purposes of face recognition was selected OpenCV library. Three different algorithms of identification were tested: FisherFaces, EigenFaces a Local Binary Patterns Histograms. Based on performance comparison best methods were implemented in developed application.

Main subjects of this thesis are texture classification and texture-based object recognition. Various texture features are being explored, including several variants of local binary patterns (LBP). A novel modification of LBP (weighted spatial LBP) is proposed, with intention to improve on the spatial coverage of the traditional LBP. Rarely used color texture features are being discussed as well. Artificial neural networks and support vector machines are used to classify all the aforementioned features. Using these methods, framework for the texture classification and image segmentation is implemented. Comprehensive texture database is employed to test its performance under different conditions. In the end, the system is applied to solve a real-world problem - the segmentation of aerial photos.

This thesis is dedicated to techniques for motion detection in video sequences. There is summary of several possible approaches to solve this matter in theoretical and practical point of view. Principles of motion detection based on difference between frames are discussed in details. Great attention is made for image description by new method Local Binary Patterns which allows creating of incredibly fast and accurate motion detector. Vital part of this thesis is application for monitoring and testing motion algorithms on video sequences in different environments. Results of measuring enables to determine best approaches for concrete place and reader might find it useful in case of designing his own motion detectors.