|
|
How to find it in the data?
Zitová, Barbara ; Šorel, Michal
The lecture aims to introduce the activities of the Image Processing Department of the Institute of Image Information of the CAS in the field of Copernicus data analysis to the professional public. The department has long been involved in the development of digital image processing and deep learning methods. During the last two years, in cooperation with the MFF UK and FJFI CTU, several student demonstration projects using data from Sentinel satellites have been finished, such as crop type recognition from Sentinel-2 time-series images, automatic segmentation of areas by land use or surface type using machine learning methods learning, more accurate cloud detection in Sentinel-2 data, in collaboration with the Institute of Hydrodynamics of the CAS Czech Republic, procedures for estimating landscape surface moisture from Sentinel-2 data and increasing the resolution of Sentinel-3 thermal data using deep learning methods. The second part will present the application of developed methods for other areas in remote sensing.
|
|
|
Multichannel blind deconvolution of color images
Brantál, František ; Šroubek, Filip (advisor) ; Šorel, Michal (referee)
The aim of the thesis is to get acquainted with methods for degradation removal (deconvolution) in scalar images in the case when no apriori information about the shape of convolution masks is known (blind deconvolution) but more than one image acquisition with different degradation is available (multichannel deconvolution). Propose possible approaches for extending blind deconvolution into multivalued (color) images using regularization forms. Implement proposed techniques and verify performance not only on synthetic data but also on real data.
|
|
|
Odstraňování odlesků z digitálních fotografií
Psotný, Dušan ; Šorel, Michal (advisor) ; Cerman, Lukáš (referee)
In this work we focus on one of the most common disturbance and that is lens flare or unwanted light scattering from light source. The main objective is to design and implement algorithms that lead to the removal of the mistakes of the image and using one or more images with diffrent exposure. Algorithms are implemented using Matlab, software program using matrices, and Image Processing Toolbox, which contains various functions for work with images. For our work, we have implemented and compared several algorithms for segmentation and removal of lens flare, we implemented an algorithm based on copying similar parts or processing of individual color channels. Our results show that we are at least partially able to delete flare.
|
|
|
Live-cell tracking in time-lapse sequences
Zámečník, Tomáš ; Šorel, Michal (advisor) ; Křivánek, Jaroslav (referee)
Title: Live-cell tracking in time-lapse sequences Author: Tomáš Zámečník Department: Department of Software and Computer Science Education Supervisor: RNDr. Michal Šorel Ph.D., Oddělení zpracování obrazu ÚTIA AV ČR Abstract: This diploma thesis deals with methods of tracking particles in image sequences. It's goal is to design and implement a complete system for tracking of live cells, their motion and division. The thesis uses conclusions of published scientific papers, studies their application and analyzes possibilities for their mo- difications or improvement. As a result, there are two applications. First of them is a demonstrational pro- gram, provided as an attachment of this thesis. Second implementation is a mo- dule of commercial software NIS-Elements, by Laboratory Imaging, Ltd., which is used by both scientific and commercial institutions in the whole world. Keywords: cell tracking, particle tracking, cell division 1
|
|
| |
|
|
Evaluation of sleep quality from infrared video
Rathan, Dominik ; Šorel, Michal (advisor) ; Zitová, Barbara (referee)
Measuring the quality of sleep is commonly realized through polysomnography, a reliable yet complicated sleep study that takes place in a sleep laboratory. The goal of this work was to develop a desktop application that would provide tools for sleep quality evaluation, based on a video recording taken in a patient's home environment using a camera with infrared vision. A motion detection algorithm based on the frame difference method was implemented, in order to assemble a graph of the patient's motion activity throughout his sleep, which was then used to approximate lengths of intervals of sleep and wakefulness and the values of basic sleep variables. It made sense to also investigate factors that could have impact on the patient's sleep quality, like the amount of light that the patient is exposed to during his sleep. Multiple algorithms for analysis of the (relative) amount of illuminance from the video recording are presented, based on modeling the lightning in the room, subspace analysis and matrix factorization. Results of these methods were verified by an experiment, and some of them can be considered reliable. 1
|
|
|
Segmentation and classification of LIDAR data
Dušek, Dominik ; Šorel, Michal (advisor) ; Obdržálek, David (referee)
The goal of this work was to design fast and simple methods for processing point-cloud-data of urban areas for virtual reality applications. For the visualization of methods, we developed a simple renderer written in C++ and HLSL. The renderer is based on DirectX 11. For point-cloud processing, we designed a method based on height-histograms for filtering ground points out of point cloud. We also proposed a parallel method for point cloud segmentation based on the region growing algorithm. The individual segments are then tested by simple rules to check if it is or it is not corresponding to a predefined object.
|
|
|
Generative neural networks in image reconstruction
Honzátko, David ; Šorel, Michal (advisor) ; Zvirinský, Peter (referee)
Recent research in generative models came up with a promising approach to modelling the prior proba- bility of natural images. The architecture of these prior models is based on deep neural networks. Although these priors were primarily designed for generating new natural-like images, its potential use is much broader. One of the possible applications is to use these models for solving the inverse problems in low-level vision (i.e., image reconstruction). This usage is mainly possible because the architecture of these models allows computing the derivative of the prior probability with respect to the input image. The main objective of this thesis is to evaluate the usage of these prior models in image reconstruction. This thesis proposes a novel model-based optimization method to two image reconstruction problems - image denoising and single-image super-resolution (SISR). The proposed method uses optimization algorithms for finding the maximum-a- posteriori probability, which is defined using the above mentioned prior models. The experimental results demonstrate that the proposed approach achieves reconstruction performance competitive with the current state-of-the-art methods, especially regarding SISR.
|
|
|
METHODOLOGY OF VOICE DISORDER EVALUATION FROM VIDEOKYMOGRAPHY RECORDS
Vydrová, J. ; Švec, J. G. ; Zitová, Barbara ; Novozámský, Adam ; Zita, Aleš ; Šorel, Michal
The aim of the methodology „Evaluating voice disorders from videkymographic data“ is to provide a comprehensive information on a new diagnostic method for diagnosis of functional and organic disorders of the voice - videokymography (VKG). The methodology is intended for otorinolaryngology and phoniatry specialists. The methodology will also serve as a study material for both trainers and trainees of the certified course „Videokymography in Medical Practice“. The diagnostic method consists of a patient examination technique and of an evaluation of the examination results. Thie data are automatically evaluated by software created to support the VKG diagnostics.
|
|
|
Image compression simultaneously in spatial and frequency domains
Vaněk, Martin ; Šorel, Michal (referee) ; Rajmic, Pavel (advisor)
This thesis is concerned with image compression. Firstly, JPEG lossy compression and origin of typical artifacts is explained. Then, some basic concepts are introduced and sparse representation of signals are presented. Then presentation of proximal algorithms is folowed and formulation of two optimazation problems, wchich removes typical artifacts on edges, created by JPEG compression. These problems are then solved by proximal algorithms.
|
guest ::
