Efficient registration of limited field of view ocular fundus imagery

Date
2013-12
Authors
Van der Westhuizen, Christo Carel
Journal Title
Journal ISSN
Volume Title
Publisher
Stellenbosch : Stellenbosch University
Abstract
ENGLISH ABSTRACT: Diabetic- and hypertensive retinopathy are two common causes of blindness that can be prevented by managing the underlying conditions. Patients suffering from these conditions are encouraged to undergo regular examinations to monitor the retina for signs of deterioration. For these routine examinations an ophthalmoscope is used. An ophthalmoscope is a relatively inexpensive device that allows an examiner to directly observe the ocular fundus (the interior back wall of the eye that contains the retina). These devices are analog and do not allow the capture of digital imagery. Fundus cameras, on the other hand, are larger devices that o er high quality digital images. They do, however, come at an increased cost and are not practical for use in the eld. In this thesis the design and implementation of a system that digitises imagery from an ophthalmoscope is discussed. The main focus is the development of software algorithms to increase the quality of the images to yield results of a quality closer to that of a fundus camera. The aim is not to match the capabilities of a fundus camera, but rather to o er a cost-e ective alternative that delivers su cient quality for use in conducting routine monitoring of the aforementioned conditions. For the digitisation the camera of a mobile phone is proposed. The camera is attached to an ophthalmoscope to record a video of an examination. Software algorithms are then developed to parse the video frames and combine those that are of better quality. For the parsing a method of rapidly selecting valid frames based on colour thresholding and spatial ltering techniques are developed. Registration is the process of determining how the selected frames t together. Spatial cross-correlation is used to register the frames. Only translational transformations are assumed between frames and the designed algorithms focuses on estimating this relative translation in a large set of frames. Methods of optimising these operations are also developed. For the combination of the frames, averaging is used to form a composite image. The results obtained are in the form of enhanced grayscale images of the fundus. These images do not match those captured with fundus cameras in terms of quality, but do show a signi cant increase when compared to the individual frames that they consists of. Collectively a set of video frames can cover a larger region of the fundus than what they do individually. By combining these frames an e ective increase in the eld of view is obtained. Due to low light exposure, the individual frames also contain signi cant noise. In the results the noise is reduced through the averaging of several frames that overlap at the same location.
AFRIKAANSE OPSOMMING: Diabetiese- en hipertensiewe retinopatie is twee algemene oorsake van blindheid wat deur middel van die behandeling van die onderliggende oorsake voorkom kan word. Pasiënte met hierdie toestande word aangemoedig om gereeld ondersoeke te ondergaan om die toestand van die retina te monitor. 'n Oftalmoskoop word gebruik vir hierdie roetine ondersoeke. 'n Oftalmoskoop is 'n relatiewe goedkoop, analoë toestel wat 'n praktisyn toelaat om die agterste interne wand van die oog the ondersoek waar die retina geleë is. Fundus kameras, aan die ander kant, is groter toestelle wat digitale beelde van 'n hoë gehalte kan neem. Dit kos egter aansienlik meer en is dus nie geskik vir gebruik in die veld nie. In hierdie tesis word die ontwerp en implementering van 'n stelsel wat beelde digitaliseer vanaf 'n oftalmoskoop ondersoek. Die fokus is op die ontwikkeling van sagteware algoritmes om die gehalte van die beelde te verhoog. Die doel is nie om die vermoëns van 'n fundus kamera te ewenaar nie, maar eerder om 'n koste-e ektiewe alternatief te lewer wat voldoende is vir gebruik in die veld tydens die roetine monitering van die bogenoemde toestande. 'n Selfoonkamera word vir die digitaliserings proses voorgestel. Die kamera word aan 'n oftalmoskoop geheg om 'n video van 'n ondersoek af te neem. Sagteware algoritmes word dan ontwikkel om die videos te ontleed en om videogrepe van goeie kwaliteit te selekteer en te kombineer. Vir die aanvanklike ontleding van die videos word kleurband drempel tegnieke voorgestel. Registrasie is die proses waarin die gekose rame bymekaar gepas word. Direkte kruiskorrelasie tegnieke word gebruik om die videogrepe te registreer. Daar word aanvaar dat die videogrepe slegs translasie tussen hulle het en die voorgestelde registrasie metodes fokus op die beraming van die relatiewe translasie van 'n groot versameling videogrepe. Vir die kombinering van die grepe, word 'n gemiddeld gebruik om 'n saamgestelde beeld te vorm. Die resultate wat verkry word, word in die vorm van verbeterde gryskleur beelde van die fundus ten toon gestel. Hierdie beelde is nie gelykstaande aan die kwaliteit van beelde wat deur 'n fundus kamera geneem is nie. Hulle toon wel 'n beduidende verbetering teenoor individuele videogrepe. Deur dat 'n groot versameling videogrepe wat gesamentlik 'n groter area van die fundus dek gekombineer word, word 'n e ektiewe verhoging van data in die area van die saamgestelde beeld verkry. As gevolg van lae lig blootstelling van die individuele grepe bevat hul beduidende ruis. In die saamgestelde beelde is die ruis aansienlik minder as gevolg van 'n groter hoeveelheid data wat gekombineer is om sodoende die ruis uit te sluit.
Description
Thesis (MScEng)-- Stellenbosch University, 2013.
Keywords
Ophthalmoscopes, Image processing -- Digital techniques, Fundus oculi, Photographic stitching, Dissertations -- Electrical and electronic engineering
Citation