|
|
Image Deblurring in Demanding Conditions
Kotera, Jan ; Šroubek, Filip (advisor) ; Portilla, Javier (referee) ; Jiřík, Radovan (referee)
Title: Image Deblurring in Demanding Conditions Author: Jan Kotera Department: Institute of Information Theory and Automation, Czech Academy of Sciences Supervisor: Doc. Ing. Filip Šroubek, Ph.D., DSc., Institute of Information Theory and Automation, Czech Academy of Sciences Abstract: Image deblurring is a computer vision task consisting of removing blur from image, the objective is to recover the sharp image corresponding to the blurred input. If the nature and shape of the blur is unknown and must be estimated from the input image, image deblurring is called blind and naturally presents a more difficult problem. This thesis focuses on two primary topics related to blind image deblurring. In the first part we work with the standard image deblurring based on the common convolution blur model and present a method of increasing robustness of the deblur- ring to phenomena violating the linear acquisition model, such as for example inten- sity clipping caused by sensor saturation in overexposed pixels. If not properly taken care of, these effects significantly decrease accuracy of the blur estimation and visual quality of the restored image. Rather than tailoring the deblurring method explicitly for each particular type of acquisition model violation we present a general approach based on flexible automatic...
|
|
|
Použití gyroskopů a akcelerometrů k doostření fotografií pořízených mobilním telefonem
Šindelář, Ondřej ; Šroubek, Filip (advisor) ; Wilkie, Alexander (referee)
Long exposure handheld photography is coupled with the problem of blurring, which is difficult to remove without additional information. The goal of this work was to utilize motion sensors contained in modern smartphones to detect exact motion track of the image sensor during the exposure and then to remove the blur from the resulting photograph according to this data. A system was proposed which performs deconvolution using a kernel from the recorded gyroscope data. An implementation on Android platform was proved on a test smartphone device.
|
|
|
Deconvolution fluorescence microscopy of yeast cells
Štec, Tomáš ; Plášek, Jaromír (advisor) ; Heřman, Petr (referee)
Title: Deconvolution fluorescence microscopy of yeast cells Author: Tomáš Štec Department: Institute of Physics of Charles University Supervisor: prof. RNDr. Jarmoír Plášek, CSc., Institute of Physics of Charles Uni- versity Abstract: Fluorescence microscopy presents an fast and cheap alternative to more advanced imaging methods like confocal and electron microscopy, even though it is subject to heavy image distortion. It is possible to recover most of the original distortion-free image using deconvolution in computer image processing. This al- lows reconstruction of 3D structure of studied objects. Deconvolution procedure of NIS Elements AR program undergoes an thorough inspection in this diploma the- sis. It is then applied on restoration of 3D structure of calcofluor stained cell wall of budding yeast Saccharomyces cerevisiae. Changes of the structure of the cell wall during cell ageing are being examined. Cell wall of aged cells shows increased surface roughness and even ruptures at the end of cell life. Keywords: fluorescence, microscopy, deconvolution, NIS Elements AR, calcofluor, yeast, cell wall, ageing
|
|
|
Studium kinematiky rozpadů top-antitopových párů v experimentu Atlas
Berta, Peter ; Kvita, Jiří (advisor) ; Soustružník, Karel (referee)
In this thesis, I deal with the measurement of the transverse momentum spectrum of the top quark produced in top-antitop pairs at the LHC at center of mass energy 7 TeV. The analysis is carried out within the ATLAS collaboration. In the single lepton decay channel, I have performed studies on the simulation which were necessary to obtain the final spectrum from real data. I describe basic event selection rules to reduce background events. I study the efficiency of top-antitop pairs reconstruction. I study the unfolding of the measured spectrum which corrects for effects caused by imperfect resolutions. At the end, I show the measured top quark transverse momentum spectrum obtained from my analysis.
|
|
|
Image Deblurring in Demanding Conditions
Kotera, Jan ; Šroubek, Filip (advisor) ; Portilla, Javier (referee) ; Jiřík, Radovan (referee)
Title: Image Deblurring in Demanding Conditions Author: Jan Kotera Department: Institute of Information Theory and Automation, Czech Academy of Sciences Supervisor: Doc. Ing. Filip Šroubek, Ph.D., DSc., Institute of Information Theory and Automation, Czech Academy of Sciences Abstract: Image deblurring is a computer vision task consisting of removing blur from image, the objective is to recover the sharp image corresponding to the blurred input. If the nature and shape of the blur is unknown and must be estimated from the input image, image deblurring is called blind and naturally presents a more difficult problem. This thesis focuses on two primary topics related to blind image deblurring. In the first part we work with the standard image deblurring based on the common convolution blur model and present a method of increasing robustness of the deblur- ring to phenomena violating the linear acquisition model, such as for example inten- sity clipping caused by sensor saturation in overexposed pixels. If not properly taken care of, these effects significantly decrease accuracy of the blur estimation and visual quality of the restored image. Rather than tailoring the deblurring method explicitly for each particular type of acquisition model violation we present a general approach based on flexible automatic...
|
|
|
Demosaicing as an ill-posed inverse problem
Mariničová, Veronika ; Šroubek, Filip (advisor) ; Hnětynková, Iveta (referee)
Color information of a scene is only recorded partially by a digital camera.Specifically, only one of the red, green, and blue color components is sampled at each pixel.The missing color values must be estimated － a process called demosaicing. Demosaicing can be solved as an individual step in the image processing pipeline. In this case, any errors and artefacts produced by this step are carried over into further steps in the image processing pipeline and are possibly magnified. Alternatively, we can try to resolve several degradations at once in a joint solution, which eliminates this effect. We present one such solution, that in addition to demosaicing, also jointly solves denoising, deconvolution, and super-resolution in the form of a convex optimization problem. We provide an overview of demosaicing methods and evaluate the results from our solution against selected existing methods.
|
|
|
Image Restoration Based on Convolutional Neural Networks
Svoboda, Pavel ; Baláž, Teodor (referee) ; Sojka, Eduard (referee) ; Zemčík, Pavel (advisor)
Tématem práce je použití konvolučních neuronových sítí pro obecnou restauraci obrazu. Ta se typicky provádí za pomoci specializovaných metod pro konkrétní typ poškození. Model konvoluční sítě zde představuje jednotný přístup, který je aplikován na dva různé typy degradace obrazu, pohybem rozmazané snímky registračních značek a artefakty vznikající vysokou kompresí. Na modely konvolučních sítí je nahlíženo ze dvou úhlů. A to jak dobře si konvoluční sítě vedou v porovnání se současnými metodami pro restauraci konkrétního typu poškození a jak velký rozsah poškození je právě jeden model ještě schopen zpracovat. Klasické metody jsou charakteristické svým úzkým zaměřením na konkrétní typ poškození. Díky své specializaci tyto metody dosahují velmi dobrých výsledků a reprezentují tak dosažené poznání v oboru. Naproti tomu je představena myšlenka jednotného přístupu, tedy mapování poškozeného obrazu přímo na restaurovaný obraz. Ta je primárně ovlivněna současným vývojem konvolučních neuronových sítí a jejich hlubokého učení v počítačovém vidění. Právě učením konvoluční sítě lze jednoduše získat model zaměřený na konkrétní typ poškození. Ten je současně nezřídka schopen pokrýt širokou škálu úrovní konkrétního poškození. V práci je představena metoda přímého mapování z rozmazaného na ostrý obraz pro restauraci pohybem rozmazaných snímků. Ta je odvozena od modelů využívaných v počítačovém vidění pro sémantickou segmentaci obrazu. V případě odstranění kompresních artefaktů je tento přístup rozšířen o specifické učení modelu a různé modifikace samotné architektury sítě. Modely konvolučních sítí v porovnání s tradičními metodami dosahují kvalitativně lepších výsledků. Zároveň se zde představené modely jednoduše vypořádají s širokým rozsahem konkrétního poškození. Ukazuje se tak, že právě modely konvolučních sítí by mohly reprezentovat jednotný přístup pro restauraci různých typů poškozeni.
|
|
|
Bioinformatical analysis of the complex multidimensional microscopy datasets
Backová, Lenka ; Černý, Jan (advisor) ; Čapek, Martin (referee)
Microscopy is embedded in the history of life sciences and vice versa. Recent advances in the field present new challenges as new revolutionary technologies arise. Sample prepa- ration, microscope operation and data analysis have become particularly demanding re- quiring specific interdisciplinary expertise. Bioimaging data analysis is computationally demanding, as microscopy technologies can easily acquire data of exceptional size, often in terabytes. Correct analysis requires computer vision knowledge, as well as knowledge of studied biological systems and last, but not least deep understanding of microscopy technology. Tools available for the analysis of the imaging data vary from open-source customizable software with a coverage of multiple tasks to a task specific proprietary software. To choose the best tools for the analysis, analysts should know their options and tasks at hand. In bioimage analysis the tasks needed to be employed depend on the desired outcome and the acquisition technology. Amongst the possible tasks to con- sider belong deconvolution, segmentation and registration. Amount of approaches and algorithms available is progressively growing, resulting in a complex field, difficult to be easily familiar with. My thesis covers different microscopy technologies with emphasis on...
|
|
|
Advanced Methods of Perfusion Analysis in MRI
Macíček, Ondřej ; Frollo, Ivan (referee) ; Mikl, Michal (referee) ; Jiřík, Radovan (advisor)
This dissertation deals with quantitative perfusion analysis of MRI contrast-enhanced image time sequences. It focuses on two so far separately used methods -- Dynamic contrast-enhanced MRI (DCE-MRI) and Dynamic susceptibility contrast MRI (DSC-MRI). The common problem of such perfusion analyses is the unreliability of perfusion parameters estimation. This penalizes usage of these unique techniques on a regular basis. The presented methods are intended to improve these drawbacks, especially the problems with quantification in DSC in case of contrast agent extravasation and instability of the deconvolution process in DCE using advanced pharmacokinetic models. There are a few approaches in literature combining DCE and DSC to estimate new parameters of the examined tissue, namely the relaxivity of the vascular and of the interstitial space. Originally, in this scheme, the 2CXM DCE model was used. Here various models for DCE analysis are tested keeping in mind the DCE-DSC combination. The ATH model was found to perform better in this setting compared to 2CXM. Finally, the ATH model was used in alternating DCE-DSC optimization algorithm and then in a truly fully simultaneous DCE-DSC. The processing was tested using simulated and in-vivo data. According to the results, the proposed simultaneous algorithm performs better in comparison with sequential DCE-DSC, unleashing full potential of perfusion analysis using MRI.
|
|
|
Deconvolution fluorescence microscopy of yeast cells
Štec, Tomáš ; Plášek, Jaromír (advisor) ; Heřman, Petr (referee)
Title: Deconvolution fluorescence microscopy of yeast cells Author: Tomáš Štec Department: Institute of Physics of Charles University Supervisor: prof. RNDr. Jarmoír Plášek, CSc., Institute of Physics of Charles Uni- versity Abstract: Fluorescence microscopy presents an fast and cheap alternative to more advanced imaging methods like confocal and electron microscopy, even though it is subject to heavy image distortion. It is possible to recover most of the original distortion-free image using deconvolution in computer image processing. This al- lows reconstruction of 3D structure of studied objects. Deconvolution procedure of NIS Elements AR program undergoes an thorough inspection in this diploma the- sis. It is then applied on restoration of 3D structure of calcofluor stained cell wall of budding yeast Saccharomyces cerevisiae. Changes of the structure of the cell wall during cell ageing are being examined. Cell wall of aged cells shows increased surface roughness and even ruptures at the end of cell life. Keywords: fluorescence, microscopy, deconvolution, NIS Elements AR, calcofluor, yeast, cell wall, ageing
|
guest ::
