Engineering Science - Theses, Dissertations, and other Required Graduate Degree Essays

Receive updates for this collection

Saliency and Tracking in Compressed Video

Date created: 
2015-06-30
Abstract: 

Visual saliency is the propensity of a part of the scene to attract attention. Computational modeling of visual saliency has become an important research problem in recent years, with applications in quality assessment, compression, object tracking, and so on. While most saliency estimation models for dynamic scenes operate on raw video, their high computational complexity is a serious drawback when it comes to practical applications. Our approach for decreasing the complexity and memory requirements is to avoid decoding the compressed bitstream as much as possible. Since most modern cameras incorporate video encoders, this paves the way for in-camera saliency estimation, which could be useful in a variety of computer vision applications. In this dissertation we present compressed-domain features that are highly indicative of saliency in natural video. Using these features, we construct two simple and effective saliency estimation models for compressed video. The proposed models have been extensively tested on two ground truth datasets using several accuracy metrics, and shown to yield considerable improvement over several state-of-the-art compressed-domain and pixel-domain saliency models. Another contribution is a tracking algorithm that also uses only compressed-domain information to isolate moving regions and estimate their trajectories. The algorithm has been tested on a number of standard sequences, and the results demonstrate its advantages over state-of-the-art for compressed-domain tracking and segmentation, with over 30% improvement in F-measure.

Document type: 
Thesis
File(s): 
Senior supervisor: 
Ivan V. Bajic
Department: 
Applied Sciences:
Thesis type: 
(Thesis) Ph.D.

Multi-hop Data Transportation with WSM Protocol in Vehicular Ad-hoc Network

Author: 
Date created: 
2015-03-24
Abstract: 

Nowadays, the wireless communication in vehicular environment is a hot topic in automotive industry. Some communication standards and protocols have been published to support the data transportation between vehicle and vehicle and between vehicle and infrastructure, which is a big challenge due to the special characteristics of vehicular ad-hoc network. In this project, we implement a hybrid stack that combines WAVE Short Message Protocol (WSMP) with a recently developed protocol – the Reactive, Density-aware Timely Dissemination protocol (REACT-DIS). The project will evaluate the performance based on its throughput, goodput and packet loss ratio by using NS3 simulation tool. Furthermore, we develop an enhanced approach that combines the REACT-DIS protocol with the limited acknowledgement to improve the performance of data dissemination in the system. The improvement of new approach also minimizes the impacting range of last hopping so that only the space near the final receiver will be impacted by the last hopping and the bandwidth of other area is not interfered. The new approach is implemented and evaluated using NS3 network simulator as well. Simulation results demonstrate that the enhanced approach outperforms the REACT-DIS protocol in an environment with low or mid density but is not ideal in high density environment.

Document type: 
Graduating extended essay / Research project
File(s): 
Senior supervisor: 
Jie Liang
Jiang Chuan Liu
Department: 
Applied Sciences:
Thesis type: 
(Project) M.Eng.

Morphometric Analysis of Ophthalmic Optical Coherence Tomography Images

Author: 
Date created: 
2015-04-14
Abstract: 

Optical coherence tomography (OCT) has become a standard in the field of ophthalmology by providing in-vivo, high-resolution, cross-sectional images of the inner structures of the eye. This dissertation outlines the construction and application of a computational pipeline for morphometric analysis of 3D retinal and peripapillary OCT images with the goal of extracting clinically meaningful information based on shape features. The images were acquired by a prototype 1060-nm swept-source OCT system and processed to correct motion and enhance the image quality. Next, anatomical features were segmented, in particular the retinal layers and structures in the optic nerve head (ONH) and laminar regions, such as the Bruch’s membrane opening (BMO) and anterior laminar boundaries. A graph-cut based, robust 3D algorithm was implemented for automated segmentation of the retinal layers. The shape of the segmented structures was measured by quantitative parameters, such as 3D thickness of the layers and dimension of the BMO. A special focus was given to establishing anatomical correspondence across multiple OCT images, from longitudinal or cross-sectional data, via registration / atlas methods. In the first approach, retinal surfaces from two OCT images were registered by a mathematical current-based deformation followed by spherical demons registration. In the second approach, retinal surfaces and corresponding signal values (ex. layer thickness) from several OCT images were jointly varied to generate a group mean template. The geometrical and signal variability among the surfaces were measured by their distances to the template serving as the common atlas. In clinical application of the pipeline, peripapillary OCT volumes of 52 myopic eyes from both normal and glaucomatous subjects were studied. Retinal layer thicknesses, shape of the BMO, bowing of Bruch’s membrane, shape of the anterior scleral canal, anterior laminar insertion, and anterior laminar surface were measured and statistically analyzed. Significant differences were observed between the normal and glaucomatous groups and demonstrated in detail the nature of the structural deformation in the glaucomatous eyes. Additionally, the results correlated the degree of myopia to several structural changes in the region, suggesting possible clues to the mechanism behind the high glaucoma susceptibility associated with advanced myopia.

Document type: 
Thesis
File(s): 
Senior supervisor: 
Faisal Beg
Marinko Sarunic
Department: 
Applied Sciences:
Thesis type: 
(Thesis) Ph.D.

Hand Tracking and its Pattern Recognition in a Network of Calibrated Cameras

Author: 
Date created: 
2015-04-01
Abstract: 

This thesis presents a vision-based approach for hand gesture recognition which combines both trajectory and hand posture recognition. The hand area is segmented by fixed-range CbCr from cluttered and moving backgrounds, and tracked by Kalman Filter. With the tracking results from two calibrated cameras, the 3D hand motion trajectory can be reconstructed. It is then modeled by dynamic movement primitives (DMP) and a support vector machine (SVM) is trained for trajectory recognition. Scale-invariant feature transform (SIFT) is employed to extract features on segmented hand postures, and a novel strategy for hand posture recognition is proposed. A gesture vector is introduced to recognize hand gesture as a whole which combines the recognition results of motion trajectory and hand postures, where an SVM is trained for gesture recognition based on gesture vectors.

Document type: 
Thesis
File(s): 
Senior supervisor: 
Shahram Payandeh
Department: 
Faculty of Applied Sciences: School of Engineering Science
Thesis type: 
(Thesis) M.A.Sc.

Multiply-and-Forward and Phase-Forward Two Way Relays

Author: 
Date created: 
2015-04-16
Abstract: 

Relay/cooperative communication has been an active area of research in the past few years. In most of the investigations, the amplifiers used in the system are assumed to be linear. In this thesis, we first investigate the effect of amplifier non-linearity has on the bit-error-rate (BER) of a two-way cooperative communication system that employs linear modulation and amplify-and-forward (AF). It was found that irreducible error floor arises, caused by the inter-modulation effect with non-linear amplification and the subsequent imperfect self-interference cancellation. As a result of this finding, we propose two signal forwarding techniques that offer MPSK modulations immunity against non-linear amplifier distortion. The first technique, termed multiply-and-forward (MF), scales the product of the two received signals at the relay for downlink transmission. The second technique, termed phase-forward (PF), has a similar product structure, except that the amplitude information in the product signal is discarded. We show that both schemes do not exhibit the irreducible error characteristic found in AF when the amplifier is non-linear. The PF scheme, in particular, can attain a BER lower than that of AF even when the amplifier is linear. The conclusion is reached that MF and PF are suitable signal forwarding strategies for two-way cooperative communication in the presence of amplifier non-linearity, with PF being the more robust of the two. Due to the multiplicative nature of the MF and PF, self-informantion cancellation becomes simple when differential modulation is used in the two-way cooperative relay system. With differential modulation, there is no explicit channel state information (CSI) estimation and and this leads to a saving of pilot symbol overhead. In the second and third parts of this thesis proposal, we extend the MF and PF protocols to DPSK and CPFSK modulations. Specifically, we demonstrate that the proposed forwarding methods has low computational complexity at the relay as well as simple self-information cancellation at the receiver in the last phase of the cooperative communication. Based on the results we have obtained on the new relay forwarding protocols, we can see that the MF and PF are two promising forwarding techniques and they worth further study and refinement.

Document type: 
Thesis
File(s): 
Senior supervisor: 
Paul Ho
Department: 
Applied Sciences:
Thesis type: 
(Thesis) Ph.D.

Very High Throughput (VHT) Multi-User Multiple Input Multiple Output (MU-MIMO) Communication in 802.11ac

Date created: 
2015-04-28
Abstract: 

Very High Throughput Multi-User Multiple Input Multiple Output (VHT MU-MIMO) is an 802.11ac communication mode that allows an Access Point (AP) to simultaneously transmit multiple data streams as Aggregated Multi-Protocol Data Units (A-MPDUs) to a group of multiple stations (STAs) over the same channel. This mode combines communication technologies that enable the 802.11ac protocol to use spectrum more efficiently compared to the previous standards. However, VHT MU-MIMO wastes an unused part of the Physical Protocol Data Unit (PPDU) interval when short and long data streams are grouped together. In this thesis, we propose a solution that improves VHT MUMIMO communications by reducing wasted portion of the PPDU duration of short data streams by concatenating longer data streams in consecutive groups. Simulations of the VHT MUMIMO communication process with and without the proposed approach indicate smaller wasted part and shorter transmission time of randomly generated STAs data streams.

Document type: 
Graduating extended essay / Research project
File(s): 
Senior supervisor: 
Ljiljana Trajkovic
Department: 
Applied Sciences:
Thesis type: 
(Project) M.Eng.

Flexible Touchpads Based on Inductive Sensors Using Conductive Composite Polymer and Flexible Metal PCB

Author: 
Date created: 
2015-03-11
Abstract: 

In this thesis, the design, fabrication, testing and characterization of two tactile sensor array technologies are presented. The two sensor systems both use inductance as a transduction principle and are designed to be implemented in flexible wearable systems. The tactile sensor arrays feature flexible PCB substrates and/or flexible conductive composite polymer (CCP) structures, resulting in highly flexible tactile arrays. Each switch consists of 4 elements: fascia, target, spacer and a sensor coil. The user presses the fascia, bringing the target in contact with the sensor coil. In the first method, where the system operates based on the principle of eddy current generation, movement of the conductive target changes the inductance of a sensing coil. In the second method, the contact between the target and the conductive membrane triggers a touch detect in the detection mechanism. An electronic circuit that accurately measures inductances is developed to detect the change in the inductance of each sensor’s coil when the user presses the target element of the sensor. Different sizes and geometries of coils in both flexible PCB metal and CCP are investigated to determine which couple best with the CCP that forms the target for the inductive coils. Using COMSOL, a COMSOL simulation of the coil geometries is also conducted in order to study the generated magnetic field and distribution of the magnetic flux density at the center of the coils. Techniques for patterning two-layer inductive coils on flexible PCBs are described. A low cost microfabrication technique to create inductive flexible coils using embedded CCP in polydimethylsiloxane as an alternative to PCB metal coils is also presented. The inductance for a sensor composed of PCB metal coils and CCP target are measured to be approximately 33.1 μH and 42.9 μH for circular and square geometries, respectively, before being pressed. When pressed, a 40% change in the inductance is observed, a change which is easily detected. In the case of coils made of CCP, although the measured inductance values are shifted because of the internal resistance of the coils, a 35% change in the inductance was observed when the conductive target was pressed.

Document type: 
Thesis
File(s): 
Senior supervisor: 
Bonnie Gray
Department: 
Applied Sciences:
Thesis type: 
(Thesis) M.A.Sc.

Designs of application-specific multiview / 3D video coding

Author: 
Date created: 
2014-11-13
Abstract: 

Many applications of multiview or three dimensional (3D) videos have been developed. This poses great challenges to video coding. We study multiview video coding (MVC), perceptual multiview video coding, 3D geometry compression, interactive multiview video streaming (IMVS), and free viewpoint video (FVV). The applications studied in this thesis can be classified into two categories. In the first category we focus on rate-distortion (RD) performance, where the distortion can be measured by mean squared errors (MSE), human visual system based MSE, or metro distance. First, we consider the application of FVV and propose a novel inpainting assisted approach to efficiently compress multiview videos. The decoder can independently recover missing data via inpainting, resulting in lower rate. Second, we study the application of just noticeable distortion (JND)-based MVC and propose to exploit inter-view or temporal redundancy of JND maps to synthesize or predict target JND maps, which are then used to adjust prediction residuals. Third, we study 3D geometry compression and propose a new 3D geometry representation. We project 3D geometry to a collection of surrounding tiles, and subsequently encode these tile images using a modified MVC. The crux of the scheme is the optimal placement of image tiles. In the second category, we study applications where real-time computation and the associated complexity also need to be considered, in addition to the RD performance. These applications include IMVS and FVV. We first consider view switching in IMVS, an application where a network client requests from server a single view at a time but can periodically switch to other views as the video is played back uninterrupted. We propose the optimal frame structure such that frequently chosen view switches are pre-computed while infrequent ones are computed in real-time upon request. On the other hand, we examine the decoder side computational complexity of view synthesis in FVV. We propose to optimally tradeoff transmission rate for decoder-side complexity. For regular view synthesis, we find the optimal subset of intermediate views to code. For a novel inpainting assisted paradigm, we argue that some expensive operations can be avoided by directly sending intra-coded blocks.

Document type: 
Thesis
File(s): 
Senior supervisor: 
Jie Liang
Department: 
Applied Sciences:
Thesis type: 
(Thesis) Ph.D.

Novel Imaging Biomarker Extraction Method for Prognostic Applications in Neuroimaging

Date created: 
2014-10-28
Abstract: 

Biomarkers derived from brain magnetic resonance imaging have promise in being able to assist in the clinical diagnosis of brain pathologies. Imaging biomarkers are a compact representation of knowledge extracted from the medical images. They can be derived from the shape of a particular brain organ, or from the deformation of a region of interest or based on the clinical understanding of a disease. In this thesis, we present novel imaging biomarkers that demonstrate potential for prognostic applications in neuroimaging and dementia. One imaging biomarker is based on the graph-theoretic analysis of inter-regional co-variation in cortical thickness, while the other is based on the Laplacian Eigen decomposition of the segmentation of a brain organ. We test these features on three distinct, but related, classification problems i.e. early detection of Alzheimer disease (AD), differential diagnosis of AD and Frontotemporal disease, and earlier detection of AD via the sub-classification of multiple domain amnestic mild cognitive impairment (MCI). We also present a novel cross- validation method that can handle class imbalance, and present comprehensive analysis into their classification performance.

Document type: 
Thesis
File(s): 
Senior supervisor: 
Mirza Faisal Beg
Department: 
Applied Sciences:
Thesis type: 
(Thesis) Ph.D.

Visual attention retargeting

Date created: 
2014-12-08
Abstract: 

This thesis explores attention retargeting---a concept related to visual saliency where the content or composition of an image is altered in an effort to guide the viewer's attention. Attention retargeting is currently in its infancy with numerous unexplored possibilities, no common methodology for evaluating performance, and no unified framework. The difficulty of attention retargeting as a saliency inversion problem lies in the lack of one-to-one mapping between saliency and the image domain, in addition to the possible negative impact of saliency alterations on image naturalness. Several approaches from recent literature to solve this challenging problem are reviewed in this context. Two novel attention retargeting methods are proposed to efficiently compute a region's propensity for drawing attention after it has been modified. The first method manipulates the orientation of a selected region, while the second modifies its hue. Both methods are applied to maximize the saliency of selected regions in various images. The likelihood of drawing attention towards the modified regions is evaluated through eye-tracking. Subjective experiments, in which participants are told to decide which image looks better between two alternatives, are used to measure the relative naturalness of the modification. An experiment was conducted to determine whether subliminal flicker is capable of drawing attention in natural images without the viewer's knowledge. Flicker was introduced to selected regions in a set of images by alternating the contrast in these regions from high to low at a frequency of 50 Hz. A comparison of eye-tracking data between participants who viewed the flickering images against those who viewed the original images suggests that subliminal flicker may, on average, repel attention rather than attract it.

Document type: 
Thesis
File(s): 
Senior supervisor: 
Ivan V. Bajic
Department: 
Applied Sciences:
Thesis type: 
(Thesis) M.A.Sc.