Theses and Dissertations from UMD
Permanent URI for this communityhttp://hdl.handle.net/1903/2
New submissions to the thesis/dissertation collections are added automatically as they are received from the Graduate School. Currently, the Graduate School deposits all theses and dissertations from a given semester after the official graduation date. This means that there may be up to a 4 month delay in the appearance of a give thesis/dissertation in DRUM
More information is available at Theses and Dissertations at University of Maryland Libraries.
Browse
3 results
Search Results
Item Deep-Learning Based Image Analysis on Resource-Constrained Systems(2021) Lee, Eung Joo; Bhattacharyya, Shuvra S; Electrical Engineering; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)In recent years, deep learning has led to high-end performance on a very wide variety of computer vision tasks. Among different types of deep neural networks, convolutional neural networks (CNNs) are extensively studied and utilized for image analysis purposes, as CNNs have the capability to effectively capture spatial and temporal dependencies in images. The growth in the amount of annotated image data and improvements in graphics processing units are factors in the rapid gain in popularity of CNN-based image analysis systems. This growth in turn motivates investigation into the application of CNN-based deep learning to increasingly complex tasks, including an increasing variety applications at the network edge. The application of deep CNNs to novel edge applications involves two major challenges. First, in many of the emerging edge-based application areas, there is a lack of sufficient training data or an uneven class balance within the datasets. Second, stringent implementation constraints --- including constraints on real-time performance, memory requirements, and energy consumption --- must be satisfied to enable practical deployment. In this thesis, we address these challenges in developing deep-CNN-based image analysis systems for deployment on resource-constrained devices at the network edge. To tackle the challenges for medical image analysis, we first propose a methodology and tool for semi-automated training dataset generation in support of robust segmentation. The framework is developed to provide robust segmentation of surgical instruments using deep learning. We then address the problem of training dataset generation for real-time object tracking using a weakly supervised learning method. In particular, we present a weakly supervised method for surgical tool tracking based on a class of hybrid sensor systems. The targeted class of systems combines electromagnetic (EM) and vision-based modalities. Furthermore, we present a new framework for assessing the quality of nonrigid multimodality image registration in real-time. With the augmented dataset, we construct a solution using various registration quality metrics that are integrated to form a single binary assessment of image registration effectiveness as either high quality or low quality. To address challenges in practical deployment, we present a deep-learning-based hyperspectral image (HSI) classification method that is designed for deployment on resource-constrained devices at the network edge. Due to the large volumes of data produced by HSI sensors, and the complexity of deep neural network (DNN) architectures, developing DNN solutions for HSI classification on resource-constrained platforms is a challenging problem. In this part of the thesis, we introduce a novel approach that integrates DNN-based image analysis with discrete cosine transform (DCT) analysis for HSI classification. In addition to medical image processing and HSI classification, a third application area that we investigate in this thesis is on-board object detection from Unmanned Aerial Vehicles (UAVs), which represents another important domain of interest for the edge-based deployment of CNN methods. In this part of the thesis, we present a novel framework for object detection using images captured from UAVs. The framework is optimized using synthetic datasets that are generated from a game engine to capture imaging scenarios that are specific to the UAV-based operating environment. Using the generated synthetic dataset, we develop new insight on the impact of different UAV-based imaging conditions on object detection performance.Item TASK SPECIFIC EVALUATION METHODOLOGY FOR CLINICAL FULL FIELD DIGITAL MAMMOGRAPHY(2012) Liu, Haimo; Kyprianou, Iacovos S; Bioengineering; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)Purpose: The purpose of this dissertation is to evaluate the image quality of clinical Full Field Digital Mammography (FFDM) systems. This is done by evaluating image acquisition performance of clinical FFDM in a comprehensive way that accounts for scatter, focal spot un-sharpness, detector blur and anti-scatter grid performance using an anthropomorphic phantom. Additionally we intend to provide a limited evaluation of the effects that image processing in clinical FFDM has in signal detectability. Methodology: We explored different strategies and a variety of mathematical model observers in order to evaluate the performance of clinical FFDM systems under different conditions. To evaluate image acquisition performance, we tested a system-model-based Hotelling observer (SMHO) model on a bench-top system using a uniform anthropomorphic phantom for an signal known exactly background known exactly (SKE/BKE) task. We then applied this concept on two clinical FFDM systems to compare their performance. In a limited study to evaluate the effects of image processing in the detectability of FFDM, we implemented the channelized Hotelling observer (CHO) model on clinically realistic images of an anatomical phantom for an SKE/BKE task. Results: Even though the two systems use different detection technologies, there was no significant difference between their image acquisition performances quantified by the Contrast-Detail (CD) curves. We applied the CHO model to investigate the image processing algorithms used in GE Senographe DS FFDM system. For the particular SKE/BKE task with rotationally symmetric signals, the image processing tends to contribute to a non-significant reduction of system detectability. Conclusion: We provided a complete description of FFDM system performance including the image acquisition chain and post-acquisition image processing. We demonstrated the simplicity and effectiveness of both the MFHO and CHO methods in a clinical setting.Item Image representation and compression via sparse solutions of systems of linear equations(2012) Nava Tudela, Alfredo; Benedetto, John J; Applied Mathematics and Scientific Computation; Digital Repository at the University of Maryland; University of Maryland (College Park, Md.)We are interested in finding sparse solutions to systems of linear equations $mathbf{A}mathbf{x} = mathbf{b}$, where $mathbf{A}$ is underdetermined and fully-ranked. In this thesis we examine an implementation of the {em orthogonal matching pursuit} (OMP) algorithm, an algorithm to find sparse solutions to equations like the one described above, and present a logic for its validation and corresponding validation protocol results. The implementation presented in this work improves on the performance reported in previously published work that used software from SparseLab. We also use and test OMP in the study of the compression properties of $mathbf{A}$ in the context of image processing. We follow the common technique of image blocking used in the JPEG and JPEG 2000 standards. We make a small modification in the stopping criteria of OMP that results in better compression ratio vs image quality as measured by the structural similarity (SSIM) and mean structural similarity (MSSIM) indices which capture perceptual image quality. This results in slightly better compression than when using the more common peak signal to noise ratio (PSNR). We study various matrices whose column vectors come from the concatenation of waveforms based on the discrete cosine transform (DCT), and the Haar wavelet. We try multiple linearization algorithms and characterize their performance with respect to compression. An introduction and brief historical review on the topics of information theory, quantization and coding, and the theory of rate-distortion leads us to compute the distortion $D$ properties of the image compression and representation approach presented in this work. A choice for a lossless encoder $gamma$ is left open for future work in order to obtain the complete characterization of the rate-distortion properties of the quantization/coding scheme proposed here. However, the analysis of natural image statistics is identified as a good design guideline for the eventual choice of $gamma$. The lossless encoder $gamma$ is to be understood under the terms of a quantizer $(alpha, gamma, beta)$ as introduced by Gray and Neuhoff.