https://www.um.edu.mt/library/oar/handle/123456789/34635 2026-04-16T00:16:30Z 2026-04-16T00:16:30Z https://www.um.edu.mt/library/oar/handle/123456789/74422 2021-04-22T12:48:32Z 2018-01-01T00:00:00Z

Title: Physically correct transparency in real-time rendering Abstract: Raster transparency remains both relevant and an open problem in computer graphics despite continuing advances in raster ray tracing and ray tracing hardware. Unlike opaque media, transparency encompasses numerous optical phenomena. Many dedicated rendering solutions exist for individual effects; however, implementing them in a single application may result in a monolithic and inefficient renderer. Furthermore, raster solutions are often lossy approximations of correctly evaluated transparency. The first problem would be eliminated by a unified raster transparency algorithm. To date, partially unified algorithms such as Phenomenological Transparency [96, 98] exist alongside other methods that are restricted to handling colour compositing and partial coverage [20, 14]. Generic or otherwise, a common subset of transparent phenomena is utilized for evaluation. Transparent emissives such as fire do not feature in these test cases: this indicates a gap in the literature despite extensive pre-existing work regarding the simulation of fire for rendering. This work seeks to contribute to these research areas by firstly investigating the improvement of the refraction approximation generated by Phenomenological Transparency with the intention of applying the method to the rendering of variable refractive fields. Secondly, it is intended to expand this investigation to include the previously unused emission parameter of the Phenomenological Transparency scattering model for transparent media. Description: M.SC.COMPUTER SCIENCE

2018-01-01T00:00:00Z https://www.um.edu.mt/library/oar/handle/123456789/40504 2020-05-17T05:07:44Z 2018-01-01T00:00:00Z

Title: Supporting memory safety through assembly transformation Abstract: In computer systems with multiple processes, memory protection is required to avoid having one process access another's memory and corrupting it. Protection can be provided either at the hardware level through specifi c components such as the memory management unit, or by using software based techniques. While on most high end systems hardware based protection is used, lower-end devices such as microcontrollers usually omit these components in order to reduce chip size, cost, and power consumption. In these cases software based approaches are used in order to protect processes from each other. Software based techniques are also commonly used between frequently communicating processes such as web browsers and their various plug-ins, even on systems where hardware based protection is available. Commonly, software based approaches involve transforming program code in order to ensure memory safety at runtime by inserting runtime memory access checks. This is typically achieved by implementing a code transformer which performs the transformations necessary for the particular approach being taken. This results in compilers that perform single specific transformations, for single instruction sets. In this dissertation we propose a generic solution which permits the scripting of different memory protection approaches, thus allowing for experimenting with new such algorithms. The proposed approach has been implemented into a proof of concept tool which uses an aspect oriented programming like language that makes it possible to instrument assembly programs at the instruction level. This facilitates the insertion of safety checks before every instruction writing data to memory for example. The language is designed such that it is instruction set agnostic making it possible to apply the same transformation for various devices. The tool has been used to implement use case tests for the MSP430 instruction set. The two major components of this project: the designed language and the prototype code transformation tool, are evaluated in terms of design, ability to express different techniques presented in related work, and overheads introduced to transformed programs. The use case resulted in runtime overheads close to those in related work, while the implementation process was much easier with the use of the designed language and prototype tool. Description: M.SC.COMPUTER SCIENCE

2018-01-01T00:00:00Z https://www.um.edu.mt/library/oar/handle/123456789/40501 2020-05-17T05:07:45Z 2018-01-01T00:00:00Z

Title: Bridging the event semantic gap Abstract: Virtual Machine Introspection (VMI) presents the opportunity to externally ob- serve a Virtual Machine (VM) by obtaining the VM's hardware events from the virtual machine monitor (hypervisor) and analysing these events to retrieve in- formation about its executing OS state. However, the main challenge for VMI consists in the difficulty in translating low-level system information into high-level executing states of the monitored VMs OS. This research demonstrates how VM software events can be leveraged to tackle the shortcomings of the current known state-of-the-art VMI system. Subsequently, a generic real-time VM monitor that translates VM software events to hardware events that can be picked up by the VMI monitor is developed. The proposed system outperforms the current state-of- the-art hardware events only approach by attempting to be more accurate in terms of generating just the events of interest using a software-to-hardware event trans- lation mechanism based on Dynamic Binary Instrumentation (DBI) as a means to bridge the event semantic gap. This novel VMI approach is evaluated in terms of its effectiveness and the performance cost incurred on the monitored VM whilst monitoring. Good results were achieved in the context of our experimental setup both in its effectiveness and its run-time overheads on the monitored VM when compared to the hardware events only (state-of-the-art) approach. Description: M.SC.COMPUTER SCIENCE

2018-01-01T00:00:00Z https://www.um.edu.mt/library/oar/handle/123456789/35413 2020-05-24T05:07:15Z 2018-01-01T00:00:00Z

Title: Super-resolution of license plates Abstract: Nowadays, image enhancement plays an essential role in crime scene investigation and the monitoring of offences on public roads. CCTV videos usually contain several distortions, including poor illumination, low-resolution, optical distortion and compression artefacts. As a consequence, specific important details may be blurred, hindering recognition of license plates. Super-resolution can be mainly divided into two categories: Single-image super-resolution and Multiple-image super-resolution. Currently, multiple-image super-resolution is the main technique which is being used to improve image quality in forensic science. However, this approach has various limitations. In this dissertation, single-image super-resolution techniques were adapted to restore license plates. This project explores the use of class-based super-resolution techniques which were successfully implemented in face and iris super-resolution. This work proposes three different approaches to class-based single-image super-resolution. The first methodology which was used is the simple least-squares method. This technique consisted of posing the super-resolution problem as a least-squares problem to minimise the reconstruction error of the images and achieve appropriate reconstruction weights. The second approach was super-resolution by Eigen transformation. This method utilises principal component analysis (PCA) to perform the super-resolution procedure while significantly reducing the dimensionality of the dictionaries which were used. Finally, a neighbour embedding method was implemented. This method uses a patch-based approach, which divides the images into a number of overlapping patches and performs the super-resolution procedure on every patch in order to obtain the best reconstruction weights for each patch. Experiments show that the first two methods presented overfitting which significantly hinders the character recognition process due to the global nature of these approaches. Thus, they are not able to adapt to different license plate structures. On the other hand, the neighbour embedding method yielded satisfactory results. This method was consistently shown to surpass the perceived quality of license plates processed through bilinear interpolation. Description: B.SC.(HONS)COMP.SCI.

2018-01-01T00:00:00Z