Eye Fundus Image Analysis for Automatic Detection of Diabetic Retinopathy

Diabetes is a rapidly increasing worldwide problem which is characterised by defective metabolism of glucose that causes long-term dysfunction and failure of various organs. The most common complication of diabetes is diabetic retinopathy (DR), which is one of the primary causes of blindness and visual impairment in adults. The rapid increase of diabetes pushes the limits of the current DR screening capabilities for which the digital imaging of the eye fundus (retinal imaging), and automatic or semi-automatic image analysis algorithms provide a potential solution. In this work, the use of colour in the detection of diabetic retinopathy is statistically studied using a supervised algorithm based on one-class classification and Gaussian mixture model estimation. The presented algorithm distinguishes a certain diabetic lesion type from all other possible objects in eye fundus images by only estimating the probability density function of that certain lesion type. For the training and ground truth estimation, the algorithm combines manual annotations of several experts for which the best practices were experimentally selected. By assessing the algorithm’s performance while conducting experiments with the colour space selection, both illuminance and colour correction, and background class information, the use of colour in the detection of diabetic retinopathy was quantitatively evaluated. Another contribution of this work is the benchmarking framework for eye fundus image analysis algorithms needed for the development of the automatic DR detection algorithms. The benchmarking framework provides guidelines on how to construct a benchmarking database that comprises true patient images, ground truth, and an evaluation protocol. The evaluation is based on the standard receiver operating characteristics analysis and it follows the medical practice in the decision making providing protocols for image- and pixel-based evaluations. During the work, two public medical image databases with ground truth were published: DIARETDB0 and DIARETDB1. The framework, DR databases and the final algorithm, are made public in the web to set the baseline results for automatic detection of diabetic retinopathy. Although deviating from the general context of the thesis, a simple and effective optic disc localisation method is presented. The optic disc localisation is discussed, since normal eye fundus structures are fundamental in the characterisation of DR.

Repeatability in automatic sorting of curved Norway spruce saw logs

Abstract

Lossless compression of hyperspectral images

Technological progress has made a huge amount of data available at increasing spatial and spectral resolutions. Therefore, the compression of hyperspectral data is an area of active research. In somefields, the original quality of a hyperspectral image cannot be compromised andin these cases, lossless compression is mandatory. The main goal of this thesisis to provide improved methods for the lossless compression of hyperspectral images. Both prediction- and transform-based methods are studied. Two kinds of prediction based methods are being studied. In the first method the spectra of a hyperspectral image are first clustered and and an optimized linear predictor is calculated for each cluster. In the second prediction method linear prediction coefficients are not fixed but are recalculated for each pixel. A parallel implementation of the above-mentioned linear prediction method is also presented. Also,two transform-based methods are being presented. Vector Quantization (VQ) was used together with a new coding of the residual image. In addition we have developed a new back end for a compression method utilizing Principal Component Analysis (PCA) and Integer Wavelet Transform (IWT). The performance of the compressionmethods are compared to that of other compression methods. The results show that the proposed linear prediction methods outperform the previous methods. In addition, a novel fast exact nearest-neighbor search method is developed. The search method is used to speed up the Linde-Buzo-Gray (LBG) clustering method.

Wavelet Filter Parameterization for Compression of Spectral Images

Vaatimus kuvatiedon tiivistämisestä on tullut entistä ilmeisemmäksi viimeisen kymmenen vuoden aikana kuvatietoon perustuvien sovellutusten myötä. Nykyisin kiinnitetään erityistä huomiota spektrikuviin, joiden tallettaminen ja siirto vaativat runsaasti levytilaa ja kaistaa. Aallokemuunnos on osoittautunut hyväksi ratkaisuksi häviöllisessä tiedontiivistämisessä. Sen toteutus alikaistakoodauksessa perustuu aallokesuodattimiin ja ongelmana on sopivan aallokesuodattimen valinta erilaisille tiivistettäville kuville. Tässä työssä esitetään katsaus tiivistysmenetelmiin, jotka perustuvat aallokemuunnokseen. Ortogonaalisten suodattimien määritys parametrisoimalla on työn painopisteenä. Työssä todetaan myös kahden erilaisen lähestymistavan samanlaisuus algebrallisten yhtälöiden avulla. Kokeellinen osa sisältää joukon testejä, joilla perustellaan parametrisoinnin tarvetta. Erilaisille kuville tarvitaan erilaisia suodattimia sekä erilaiset tiivistyskertoimet saavutetaan eri suodattimilla. Lopuksi toteutetaan spektrikuvien tiivistys aallokemuunnoksen avulla.

Classification-Based compression of Multispectral Images

The purpose of this thesis is to present a new approach to the lossy compression of multispectral images. Proposed algorithm is based on combination of quantization and clustering. Clustering was investigated for compression of the spatial dimension and the vector quantization was applied for spectral dimension compression. Presenting algo¬rithms proposes to compress multispectral images in two stages. During the first stage we define the classes' etalons, another words to each uniform areas are located inside the image the number of class is given. And if there are the pixels are not yet assigned to some of the clusters then it doing during the second; pass and assign to the closest eta¬lons. Finally a compressed image is represented with a flat index image pointing to a codebook with etalons. The decompression stage is instant too. The proposed method described in this paper has been tested on different satellite multispectral images from different resources. The numerical results and illustrative examples of the method are represented too.

Statistical Analysis of Spectral Images

Tämä diplomityö liittyy Spektrikuvien tutkimiseen tilastollisen kuvamallin näkökulmasta. Diplomityön ensimmäisessä osassa tarkastellaan tilastollisten parametrien jakaumien vaikutusta väreihin ja korostumiin erilaisissa valaistusolosuhteissa. Havaittiin, että tilastollisten parametrien väliset suhteet eivät riipu valaistusolosuhteista, mutta riippuvat kuvan häiriöttömyydestä. Ilmeni myös, että korkea huipukkuus saattaa aiheutua värikylläisyydestä. Lisäksi työssä kehitettiin tilastolliseen spektrimalliin perustuvaa tekstuurinyhdistämisalgoritmia. Sillä saavutettiin hyviä tuloksia, kun tilastollisten parametrien väliset riippuvuussuhteet olivat voimassa. Työn toisessa osassa erilaisia spektrikuvia tutkittiin käyttäen itsenäistä komponenttien analyysia (ICA). Seuraavia itsenäiseen komponenttien analyysiin tarkoitettuja algoritmia tarkasteltiin: JADE, kiinteän pisteen ICA ja momenttikeskeinen ICA. Tutkimuksissa painotettiin erottelun laatua. Paras erottelu saavutettiin JADE- algoritmilla, joskin erot muiden algoritmien välillä eivät olleet merkittäviä. Algoritmi jakoi kuvan kahteen itsenäiseen, joko korostuneeseen ja korostumattomaan tai kromaattiseen ja akromaattiseen, komponenttiin. Lopuksi pohditaan huipukkuuden suhdetta kuvan ominaisuuksiin, kuten korostuneisuuteen ja värikylläisyyteen. Työn viimeisessä osassa ehdotetaan mahdollisia jatkotutkimuskohteita.

Computing the persistent homology of range images with alpha shapes

Laser scanning is becoming an increasingly popular method for measuring 3D objects in industrial design. Laser scanners produce a cloud of 3D points. For CAD software to be able to use such data, however, this point cloud needs to be turned into a vector format. A popular way to do this is to triangulate the assumed surface of the point cloud using alpha shapes. Alpha shapes start from the convex hull of the point cloud and gradually refine it towards the true surface of the object. Often it is nontrivial to decide when to stop this refinement. One criterion for this is to do so when the homology of the object stops changing. This is known as the persistent homology of the object. The goal of this thesis is to develop a way to compute the homology of a given point cloud when processed with alpha shapes, and to infer from it when the persistent homology has been achieved. Practically, the computation of such a characteristic of the target might be applied to power line tower span analysis.

Statistical segmentation methods and color variance analysis of retinal images

In this research, the effectiveness of Naive Bayes and Gaussian Mixture Models classifiers on segmenting exudates in retinal images is studied and the results are evaluated with metrics commonly used in medical imaging. Also, a color variation analysis of retinal images is carried out to find how effectively can retinal images be segmented using only the color information of the pixels.

Automatic recognition of fish from video sequences

The problem of automatic recognition of the fish from the video sequences is discussed in this Master’s Thesis. This is a very urgent issue for many organizations engaged in fish farming in Finland and Russia because the process of automation control and counting of individual species is turning point in the industry. The difficulties and the specific features of the problem have been identified in order to find a solution and propose some recommendations for the components of the automated fish recognition system. Methods such as background subtraction, Kalman filtering and Viola-Jones method were implemented during this work for detection, tracking and estimation of fish parameters. Both the results of the experiments and the choice of the appropriate methods strongly depend on the quality and the type of a video which is used as an input data. Practical experiments have demonstrated that not all methods can produce good results for real data, whereas on synthetic data they operate satisfactorily.

Automatic image-based identification of Saimaa ringed seals

The Saimaa ringed seal is one of the most endangered seals in the world. It is a symbol of Lake Saimaa and a lot of effort have been applied to save it. Traditional methods of seal monitoring include capturing the animals and installing sensors on their bodies. These invasive methods for identifying can be painful and affect the behavior of the animals. Automatic identification of seals using computer vision provides a more humane method for the monitoring. This Master's thesis focuses on automatic image-based identification of the Saimaa ringed seals. This consists of detection and segmentation of a seal in an image, analysis of its ring patterns, and identification of the detected seal based on the features of the ring patterns. The proposed algorithm is evaluated with a dataset of 131 individual seals. Based on the experiments with 363 images, 81\% of the images were successfully segmented automatically. Furthermore, a new approach for interactive identification of Saimaa ringed seals is proposed. The results of this research are a starting point for future research in the topic of seal photo-identification.

Images of the future of privacy: A privacy dynamics framework and a causal layered

The future of privacy in the information age is a highly debated topic. In particular, new and emerging technologies such as ICTs and cognitive technologies are seen as threats to privacy. This thesis explores images of the future of privacy among non-experts within the time frame from the present until the year 2050. The aims of the study are to conceptualise privacy as a social and dynamic phenomenon, to understand how privacy is conceptualised among citizens and to analyse ideal-typical images of the future of privacy using the causal layered analysis method. The theoretical background of the thesis combines critical futures studies and critical realism, and the empirical material is drawn from three focus group sessions held in spring 2012 as part of the PRACTIS project. From a critical realist perspective, privacy is conceptualised as a social institution which creates and maintains boundaries between normative circles and preserves the social freedom of individuals. Privacy changes when actors with particular interests engage in technology-enabled practices which challenge current privacy norms. The thesis adopts a position of technological realism as opposed to determinism or neutralism. In the empirical part, the focus group participants are divided into four clusters based on differences in privacy conceptions and perceived threats and solutions. The clusters are fundamentalists, pragmatists, individualists and collectivists. Correspondingly, four ideal-typical images of the future are composed: ‘drift to low privacy’, ‘continuity and benign evolution’, ‘privatised privacy and an uncertain future’, and ‘responsible future or moral decline’. The images are analysed using the four layers of causal layered analysis: litany, system, worldview and myth. Each image has its strengths and weaknesses. The individualistic images tend to be fatalistic in character while the collectivistic images are somewhat utopian. In addition, the images have two common weaknesses: lack of recognition of ongoing developments and simplistic conceptions of privacy based on a dichotomy between the individual and society. The thesis argues for a dialectical understanding of futures as present images of the future and as outcomes of real processes and mechanisms. The first steps in promoting desirable futures are the awareness of privacy as a social institution, the awareness of current images of the future, including their assumptions and weaknesses, and an attitude of responsibility where futures are seen as the consequences of present choices.

On the Condition Monitoring of Induction machines

The research of condition monitoring of electric motors has been wide for several decades. The research and development at universities and in industry has provided means for the predictive condition monitoring. Many different devices and systems are developed and are widely used in industry, transportation and in civil engineering. In addition, many methods are developed and reported in scientific arenas in order to improve existing methods for the automatic analysis of faults. The methods, however, are not widely used as a part of condition monitoring systems. The main reasons are, firstly, that many methods are presented in scientific papers but their performance in different conditions is not evaluated, secondly, the methods include parameters that are so case specific that the implementation of a systemusing such methods would be far from straightforward. In this thesis, some of these methods are evaluated theoretically and tested with simulations and with a drive in a laboratory. A new automatic analysis method for the bearing fault detection is introduced. In the first part of this work the generation of the bearing fault originating signal is explained and its influence into the stator current is concerned with qualitative and quantitative estimation. The verification of the feasibility of the stator current measurement as a bearing fault indicatoris experimentally tested with the running 15 kW induction motor. The second part of this work concentrates on the bearing fault analysis using the vibration measurement signal. The performance of the micromachined silicon accelerometer chip in conjunction with the envelope spectrum analysis of the cyclic bearing faultis experimentally tested. Furthermore, different methods for the creation of feature extractors for the bearing fault classification are researched and an automatic fault classifier using multivariate statistical discrimination and fuzzy logic is introduced. It is often important that the on-line condition monitoring system is integrated with the industrial communications infrastructure. Two types of a sensor solutions are tested in the thesis: the first one is a sensor withcalculation capacity for example for the production of the envelope spectra; the other one can collect the measurement data in memory and another device can read the data via field bus. The data communications requirements highly depend onthe type of the sensor solution selected. If the data is already analysed in the sensor the data communications are needed only for the results but in the other case, all measurement data need to be transferred. The complexity of the classification method can be great if the data is analysed at the management level computer, but if the analysis is made in sensor itself, the analyses must be simple due to the restricted calculation and memory capacity.

Lossless Compression Methods for Multispectral Images: Development and Comparisons

Main purpose of this thesis is to introduce a new lossless compression algorithm for multispectral images. Proposed algorithm is based on reducing the band ordering problem to the problem of finding a minimum spanning tree in a weighted directed graph, where set of the graph vertices corresponds to multispectral image bands and the arcs’ weights have been computed using a newly invented adaptive linear prediction model. The adaptive prediction model is an extended unification of 2–and 4–neighbour pixel context linear prediction schemes. The algorithm provides individual prediction of each image band using the optimal prediction scheme, defined by the adaptive prediction model and the optimal predicting band suggested by minimum spanning tree. Its efficiency has been compared with respect to the best lossless compression algorithms for multispectral images. Three recently invented algorithms have been considered. Numerical results produced by these algorithms allow concluding that adaptive prediction based algorithm is the best one for lossless compression of multispectral images. Real multispectral data captured from an airplane have been used for the testing.

Prediction in transformed images

The study is related to lossless compression of greyscale images. The goal of the study was to combine two techniques of lossless image compression, i.e. Integer Wavelet Transform and Differential Pulse Code Modulation to attain better compression ratio. This is an experimental study, where we implemented Integer Wavelet Transform, Differential Pulse Code Modulation and an optimized predictor model using Genetic Algorithm. This study gives encouraging results for greyscale images. We achieved a better compression ration in term of entropy for experiments involving quadrant of transformed image and using optimized predictor coefficients from Genetic Algorithm. In an other set of experiments involving whole image, results are encouraging and opens up many areas for further research work like implementing Integer Wavelet Transform on multiple levels and finding optimized predictor at local levels.