UCL Discovery
UCL home » Library Services » Electronic resources » UCL Discovery

Image fusion: Direct application to low light level and thermal images, and efforts toward an automated medical image registration algorithm

Oakley, Jonathan; (1997) Image fusion: Direct application to low light level and thermal images, and efforts toward an automated medical image registration algorithm. Masters thesis (M.Phil), UCL (University College London). Green open access

[thumbnail of Image_fusion_Direct_applicati.pdf] Text
Image_fusion_Direct_applicati.pdf

Download (19MB)

Abstract

We begin the first half of this report with an investigation into a quite specific application of image fusion. Among the characteristics of the application were that it used two specific modalities, with some possibility of their being a third. Also, that it should operate on dynamically changing image sequences, and do so in real-time. Through the reviewing of the possible implementation methods to hand, we construct our argument for the use of one particularly promising method. This favoured a pyramidal image fusion scheme, which, having been implemented, allowed for some closing conclusions to be drawn. The method of constructing the pyramidal representation of individual images was shown to be the key element in determining the effectiveness of the fusion process. As such, it is this process that therefore commands most of the project's attention. Arguments for and against the different kernels that are used to 'build' the pyramid come from purely mathematical considerations, and naturally via some computational model that attempts to model the human visual system. Our aim is to review these and then provide some insight into how a fusion scheme might be implemented given the application. Therefore its underlying theoretical and implementation details are described ahead of the results that show its effectiveness. The general review is finally reduced to the simple yet valid tests of the fusion technique in respect of various other pixel fusion methods. Having concluded it to be the method of choice, scope now remains concerning the optimal method for performing this process, where influence has been generously afforded to the analogous human visual system. It is in the field of Medical Imaging that an increasing number of relevant imaging modalities have been made available to clinicians. Different modalities often provide complementary information; for example, some are better at showing anatomy, others physiology. As this choice and availability of image modalities increases, prospects of image correlation and fusion arise. And as such, it is believed that some combination of these images will prove requisite to improved clinical diagnosis and treatment. This project is to do with the combination and the subsequent visualisation of the different multimodality images. The work involves the registration, or alignment, of the different image data sets. The registration technique employed attempts to include some notion of intelligence by incorporating anatomical knowledge and knowledge of each imager's characteristics. Work will then require the final visualisation of the combined image sets, where, ultimately, the success of any strategy will be dependent on the success of the registration process. The layout of this report reflects this partitioning of my efforts. The first sections set its context with the literature review. Because of its importance, this review aims to be comprehensive and I make no excuses for volume; I have put a great deal of effort into the explanations given, and the knowledgeable reader may skip this. The content focuses on image registration as a prerequisite of the fusion and visualisation scheme. We describe our methods in respect of those in existence, and our results section shows the successes and failings of each. This is summarised in chapter 9, which details the conclusions of the work with a realistic look at what has been achieved. It is in the light of this discussion and the results that we derive a number of possible applications where our methods might be beneficial. These are developed in my final chapter which ends with a number of interesting and promising topics of further work. These no longer remain solely in the bounds of registration techniques, but are envisaged as natural extensions to the work of this project such that its importance is substantial and its inclusion relevant. Finally, to keep the registration work within the context of the overall document, the nature of the visualisation scheme is returned to, but only as an appendix in the back of the report. It is written in respect of the overall theme of Image Fusion and the registration work done.

Type: Thesis (Masters)
Qualification: M.Phil
Title: Image fusion: Direct application to low light level and thermal images, and efforts toward an automated medical image registration algorithm
Open access status: An open access version is available from UCL Discovery
Language: English
Additional information: Thesis digitised by ProQuest.
Keywords: Health and environmental sciences
URI: https://discovery.ucl.ac.uk/id/eprint/10102683
Downloads since deposit
40Downloads
Download activity - last month
Download activity - last 12 months
Downloads by country - last 12 months

Archive Staff Only

View Item View Item