Doctor of Philosophy (PhD), Wright State University, 2014, Computer Science and Engineering PhD

To construct a complete representation of a scene with environmental obstacles such as fog, smoke, darkness, or textural homogeneity, multisensor video streams captured in diferent modalities are considered. A computational method for automatically fusing multimodal image streams into a highly informative and unified stream is proposed. The method consists of the following steps: 1. Image registration is performed to align video frames in the visible band over time, adapting to the nonplanarity of the scene by automatically subdividing the image domain into regions approximating planar patches 2. Wavelet coefficients are computed for each of the input frames in each modality 3. Corresponding regions and points are compared using spatial and temporal information across various scales 4. Decision rules based on the results of multimodal image analysis are used to combine the wavelet coefficients from different modalities 5. The combined wavelet coefficients are inverted to produce an output frame containing useful information gathered from the available modalities Experiments show that the proposed system is capable of producing fused output containing the characteristics of color visible-spectrum imagery while adding information exclusive to infrared imagery, with attractive visual and informational properties.

Doctor of Philosophy (Ph.D.), University of Dayton, 2024, Computer Science

Image is used everywhere since it conveys a story, a fact, or an imagination without any words. Thus, it can substitute the sentences because the human brain can extract the knowledge from images faster than words. However, creating an image from scratch is not only time-consuming, but also a tedious task that requires skills. Creating an image is not a trivial task since it contains rich features and fine-grained details, such as colors, brightness, saturation, luminance, texture, shadow, and so on. Thus, in order to generate an image in less time and without any artistic skills, sketch-to-image synthesis can be used. The reason is that hand sketches are much easier to produce, where only the key structural information is contained. Moreover, it can be drawn without skills and in less time. In fact, since sketches are often simple and rough black and white and sometimes imperfect, converting a sketch into an image is not a trivial problem. Hence, it has attracted the researchers' attention to solve this challenging problem; therefore, much research has been conducted in this field to generate photorealistic images. However, the generated images still suffer from issues, such as the un-naturality, the ambiguity, the distortion, and most importantly, the difficulty in generating images from complex input with multiple objects. Most of these problems are due to converting a sketch into an image directly in one-shot. To this end, in this dissertation, we propose a new framework that divides the problem into sub-problems, leading to generating high-quality photorealistic images even with complicated sketches. Instead of directly mapping the input sketch into an image, we map the sketch into an intermediate result, namely, mask map, through an instance segmentation and semantic segmentation in two levels: background segmentation and foreground segmentation. Background segmentation is formed based on the context of the existing foreground objects. Various natural scenes a (open full item for complete abstract)

Master of Science in Computer Engineering, University of Dayton, 2023, Electrical and Computer Engineering

Deep Convolutional Neural Networks demonstrate state-of-art performance in computer vision and medical image tasks. However, handling a large-scale image is still a challenging task that usually deals with resizing and patching methods to embed in the lower dimensional space. Recently, Learnable Resizer (LR) has been proposed to analyze large-scale images for computer vision tasks. This study proposes two DCNN models for classification and segmentation tasks constructed with LR in combination with successful classification and segmentation architectures. The performance of the proposed models is evaluated for the Diabetic Retinopathy (DR) analysis and skin cancer segmentation tasks. The proposed model demonstrated better performance than the existing methods for segmentation and classification tasks. For classification tasks, the proposed architectures achieved a 5.34% improvement in accuracy compared to ResNet50. Besides, around 0.62% accuracy over the base model and 0.28% in Intersection-over-Union (IoU) from state-of-the-art performance. The proposed model with the resizer network enhances the capability of the existing R2U-Net for medical image segmentation tasks. Moreover, the proposed methods enable a significant advantage in learning better with a few samples. The experimental results reveal that the proposed models are better than the current approaches.

Master of Fine Arts, The Ohio State University, 2021, Art

Monsters are powerful symbols of transformative agency, heavily ingrained in Western culture. With transmutating creatures living rent-free in our collective imagination, I have to wonder: why is it taboo for queer people to transform? Tracing a historical line from biblical angels, Greek mythology, the gothic novel, and contemporary horror cinema, I create a framework for understanding monsters as revered, transformative figures in important texts throughout the centuries. Just as LGBTQ+ activists reclaimed `queer' as a radical identifier, I reclaim `monster' as an uncompromising symbol of bodily agency, engaging with Queer readings and critical media theory along the way. Using my MFA Thesis artwork God Made Me (And They Love Me), I weave my soft sculpture beasties through historical imagery, religious text, folklore, and media pieces depicting `monster' and `monstrosity'.

Master of Science (M.S.), University of Dayton, 2015, Electrical Engineering

Image quality has always been an important aspect of the image processing field. Subjective quality is useful since images are a visual medium, but objective quality measures are needed because they are unbiased and can be used as parts of larger processing systems. Many image quality metrics exist that attempt to give an objective score to an image based on its likeness to a reference. These metrics work well if the reference is known and the test image is assumed to be a distorted version of the reference. However, in areas such as image enhancement, the reference image is generally worse than the test image and measuring likeness between the two is not a good indication of visual quality. A no-reference image enhancement quality metric is proposed in this paper that uses three factors to score images: lightness, contrast, and noise. It has been shown in literature that certain ideal ranges for lightness and contrast exist, and image enhancement techniques tend to push an image towards these. The metric gives each pixel in an image a score based on its neighborhood statistics. An image fusion technique is also proposed that fuses multiple enhanced images into one based on the local scores obtained from the no-reference metric. It is shown that this fused image scores higher using the no-reference metric and also has superior visual quality.

Master of Science (MS), Wright State University, 2013, Human Factors and Industrial/Organizational Psychology MS

This study tested the hypothesis that the similarity of the cue and target in a visual search task is related to performance. Specifically, it was hypothesized that as the similarity between the cue and the target along the dimensions of stimulus contrast, spatial resolution and size increases, the amount of time that it takes to find a target among distractors decreases. Three experiments were performed to investigate the question. Experiments 1 and 2 employed a methodology that employed homogeneous search arrays where the contrast, spatial resolution and size of the elements were constant (high contrast, high spatial resolution and large size) and resulted in two small, statistically significant size effects. Experiment 3 was designed with heterogeneous search arrays for the task. This redesign produced larger performance differences that supported the similarity hypothesis. Differences in size produced the largest performance shifts, followed by differences in spatial resolution and differences in contrast producing smaller effects.

Master of Science (M.S.), University of Dayton, 2024, Electrical Engineering

RGB imagery provides detail which is usually sufficient to perform computer vision tasks. However, images taken in low-light appear vastly different from well-lit imagery due to the diversity in light intensity. Polarimetric data provides additional detail which focuses on the orientation of the light rather than intensity. Scaling our classic RGB images using polarimetric data can maintain the RGB image type, while also enhancing image contrast. This allows transfer learning using pre-trained RGB models to appear more feasible. Our work focuses on developing a large dataset of paired polarimetric RGB images in a highly controlled laboratory environment. Then, we perform transfer learning on a pre-trained image segmentation model with each of our image product types. Finally, we compare these results in both well-lit and low-light scenarios to see how our polarimetrically enhanced RGB images stack up against regular RGB images.

Doctor of Philosophy, The Ohio State University, 2023, Nuclear Engineering

Neutron radiography and computed tomography (CT) offer unique opportunities in the field of non-destructive evaluation (NDE) for both reactor-based and accelerator-based neutron sources. The most widely implemented and advanced state-of-the-art techniques in radiography and CT use X-ray sources. However, in scenarios where X-ray penetration or contrast among materials are limited, advanced thermal and fast neutron methods can offer additional insights. X-ray attenuation is generally minimal for materials with a low atomic number (Z) and increases as the atomic number grows. This characteristic can sometimes result in inadequate contrast for low-Z materials or excessive attenuation for high-Z materials. Thermal neutron radiography – using neutrons near the thermal equivalent energy of 0.025 eV – takes advantage of the variable thermal neutron capture cross sections as a function of Z to provide high contrast, especially for certain elements such as Li and B. Attenuation values for thermal neutrons do not tend to follow a specific trend which enables contrast to be high for specific combinations of elements or isotopes. However, these lower energy neutrons have difficulty penetrating thicker objects, and certain elements can become activated due to nuclear transformations. Fast neutron radiography (using ~MeV neutrons) exhibits a more consistent attenuation across varying Z values, potentially benefiting both low-Z and high-Z materials. The primary advantage of fast neutron radiography and CT lies in the ability of MeV neutrons to penetrate high-Z materials like lead and tungsten better than MeV X-rays. The main challenge is ensuring adequate contrast between materials and achieving high detection efficiency, since these high energy neutrons are so penetrating. Generally, the most probable interactions of elements with fast neutrons are elastic scattering interactions which tends to reduce activation of target materials compared to thermal neutron techniques. This can (open full item for complete abstract)

Master of Computer Science (M.C.S.), University of Dayton, 2022, Computer Science

A picture is worth a thousand words. A single image has the power to influence individuals and change their behaviour, whereas a single word does not. Even a barely visible image, displayed on a screen for only a few milliseconds, appears to be capable of changing one's behaviour. In this thesis, we experimentally investigated the relationship between facial expressions and perceived emotions. To this end, we built two datasets, namely, the image dataset for image emotion analysis and the face dataset for expression recognition. During the annotation of the image dataset, both facial expressions and perceived emotions are recorded via a mobile application. We then use a classifier trained on the face dataset to recognize the user's expression and compare it with the perceived emotion.

Master of Science in Electrical Engineering, University of Dayton, 2022, Electrical and Computer Engineering

With current sensor technology, visible wavelength (VIS) images can be acquired at very high resolutions (HR) compared to the infrared (IR) images. Therefore, image fusion techniques aim to augment IR images with the superior spatial resolution of VIS images to overcome the resolution problems in IR imaging. This thesis introduces two ways to integrate IR and VIS images, IR image super-resolution and IR and VIS image fusion. The first application is super-resolution (SR) for IR images. We propose an IR image SR algorithm based on U-Net. By fusing the HR image features of the VIS images, the network can produce an IR SR image successfully and efficiently. Secondly, we also propose a novel framework for combining VIS and IR images, guided by feature extraction techniques such as VGG16. By designing the algorithm to preserve the meaningful VGG16 features from both IR and VIS images, the proposed method achieves excellent performance in the qualitative and quantitative aspects. In addition, we propose joint super-resolution and image fusion between IR and VIS images. Finally, we developed a new HR VIS and LR IR image pair dataset. Since this data collection closely resembles the real-world sensing scenarios, it is a valuable resource for continued exploration of this image processing field.

Master of Computer Science (M.C.S.), University of Dayton, 2020, Computer Science

Text-to-image is a process of generating an image from the input text. It has a variety of applications in art generation, computer-aided design, and photo-editing. In this thesis, we propose a new framework that leverages mask anchor points to incorporate two major steps in the image synthesis. In the first step, the mask image is generated from the input text and the mask dataset. In the second step, the mask image is fed into the state-of-the-art mask-to-image generator. Note that the mask image captures the semantic information and the location relationship via the anchor points. We develop a user-friendly interface that helps parse the input text into the meaningful semantic objects. However, to synthesize an appealing image from the text, image aesthetics criteria should be considered. Therefore, we further improve our proposed framework by incorporating the aesthetic assessment from photography composition rules. To this end, we randomize a set of mask maps from the input text via the anchor point-based mask map generator, and then we compute and rank the image aesthetics score for all generated mask maps following two composition rules, namely, the rule of thirds along with the rule of formal balance. In the next stage, we feed the subset of the mask maps, which are the highest, lowest, and the average aesthetic scores, into the state-of-the-art mask-to-image generator via image generator. The photorealistic images are further re-ranked to obtain the synthesized image with the highest aesthetic score. Thus, to overcome the state-of-the-arts generated images' problems such as the un-naturality, the ambiguity, and the distortion, we propose a new framework. Our framework maintains the clarity of the entities' shape, the details of the entity edges, and the proper layout no matter how complex the input text is and how many entities and spatial relations in the text. Our contribution is converting the input text to an appropriate constructed mask map or to a set (open full item for complete abstract)

Doctor of Philosophy (PhD), Ohio University, 2018, Electrical Engineering & Computer Science (Engineering and Technology)

Changes in the morphology of a skin lesion is indicative of melanoma, a deadly type of skin cancer. This dissertation proposes a temporal analysis method to monitor the vascularity, pigmentation, size and other critical morphological attributes of the lesion. Digital images of a skin lesion acquired during follow-up imaging sessions are input to the proposed system. The images are pre-processed to normalize variations introduced over time. The vascularity is modelled as the skin images' red channel information and its changes by the Kullback-Leibler (KL) divergence of the probability density function approximation of histograms. The pigmentation is quantified as textural energy, changes in the energy and pigment coverage in the lesion. An optical flow field and divergence measure indicates the magnitude and direction of global changes in the lesion. Sub-surface change is predicted based on the surface skin lesion image with a novel approach. Changes in key morphological features such as lesions' shape, color, texture, size, and border regularity are computed. Future trends of the skin lesions features are estimated by an auto-regressive predictor. Finally, the features extracted using deep convolutional neural networks and the hand-crafted lesion features are compared with classification metrics. An accuracy of 80.5%, specificity of 98.14%, sensitivity of 76.9% with a deep learning neural network is achieved. Experimental results show the potential of the proposed method to monitor a skin lesion in real-time during routine skin exams.

Master of Science in Computer Engineering (MSCE), Wright State University, 2018, Computer Engineering

Due to current market forces, leading-edge semiconductor fabrication plants have moved outside of the US. While this is not a problem at first glance, when it comes to security-sensitive applications, over-production, device cloning, or design alteration becomes a possibility. Since these vulnerabilities exist during the fabrication phase, a Reverse Engineering (RE) step must be introduced to help ensure secure device operation. This thesis proposes several unique methods and a collection of tools to ensure trust assurance in integrated circuit design by detecting fabrication flaws and possible hardware Trojans using several image processing techniques; fused into a singular view of the design. This suite of tools, the Automated Iterative Reverse Engineer (AIRE), addresses these security concerns. AIRE is comprised of an Image Stitcher, Stitch Assembler, Design Parser, Image Matcher, and IC Image Fusion Viewer. These tools work in concert to assist the design validation test engineers in assessing and analyzing the structure and performance of the IC. The first tool, the Image Stitcher, is the focus of this thesis.

Master of Science (MS), Wright State University, 2016, Computer Science

Many video processing applications, such as motion detection and tracking, rely on accurate and robust alignment between consecutive video frames. Traditional approaches to video image registration, such as pyramidal Kanade-Lucas-Tomasi (KLT) feature detection and tracking are fast and subpixel accurate, but are not robust to large inter-frame displacements due to rotation, scale, or translation. This thesis presents an alternative hybrid approach using normalized gradient correlation (NGC) in the frequency domain and normalized cross-correlation (NCC) in the spatial domain that is fast, accurate, and robust to large displacements. A scale space search is incorporated into NGC to enable more consistent recovery of scale factors up to 6. Results show that the scale space enhanced NGC improves performance in both speed and maximum scale recovery. The proposed hybrid approach is compared to KLT and results demonstrate a significant improvement in robustness in exchange for a slight reduction in accuracy.

Master of Arts (M.A.), Xavier University, 2016, Psychology

The purpose of this research was to determine if participation in a workplace health promotion (WHP) program has any influence on employees' awareness of and satisfaction with their body image, and whether or not those levels of awareness and satisfaction differ between male and female employees. It was hypothesized that individuals who participate in a WHP program would report higher body surveillance and lower body shame than those who do not participate in such a program. Exploratory hypotheses also questioned if there were gender differences in participants' experiences of surveillance and shame. Data was collected from a sample of 174 participants. Independent-samples t-tests were used to test the main hypotheses focused on participation, and ANCOVAs were used to test the exploratory hypotheses focused on gender. Neither of the main hypotheses yielded significant results, whereas the exploratory hypotheses yielded results in the opposite direction of what was hypothesized. This study contributed interesting findings to the literature on WHP and wellness programs, as well as two keys factors that contribute to the development of body image. Participating in, or at least being exposed to, a WHP program may lead some employees to experience varying degrees of body surveillance or body shame, but said participation did not yield a significant increase in surveillance nor a significant decrease in shame. Of greater interest is that although male WHP program participants and non-participants alike reported significantly more body surveillance, there were no significant differences between both male and female participants (and non-participants) in the experience of body shame. These results have significant implications for future discussions surrounding the development of body image and experience in WHP programs.

Master of Science (M.S.), University of Dayton, 2016, Electrical Engineering

This study suggests improvements and an extension for the No-Reference Image Enhancement Quality Metric And Fusion (NRIE-QMF) Technique, that measures a perceptual quality score. To mesure the quality score, the NRIE-QMF metric uses the image statistics based on brightness, contrast, and noise content. The NRIE-QMF uses several image inputs from various image enhancement methods (GHE, CLAHE, and LTSN) and calculates a score value for each pixels based on the local neighborhood statistics. Then respective pixel scores of each enhanced image are weighted and fused into one to create a combined image. The NRIE-QMF metric is analyzed for execution time using the MATLAB profiler. Few modification and optimization steps are carried out to increase the execution speed while maintaining a good output. Secondly, enhanced images are scored using the proposed metric and the score matrices are thresholded compared to the original image's score matrix to avoid over-amplification caused by some enhancement methods. Finally, it is shown that the proposed metric achieves a 85.8% speed increase compared to the NRIE-QMF method and generates a combined output image with a superior visual quality. Also, quality score of the new combined image results higher than those of the enhanced images used for fusion, demonstrating the superiority of the proposed method's fusion technique.

Doctor of Philosophy (Ph.D.), University of Dayton, 2015, Electrical Engineering

Multi-frame super-resolution (SR) image reconstruction methods seek to overcome sampling limitations in staring cameras, leading to the reduction or elimination of aliasing artifacts in imaging systems. Prior research has predominantly been focused on single band or panchromatic imaging applications. Sampling limitations are further exacerbated for cameras that add functionality by implementing spatial filters across the cameras focal plane array. The types of cameras, generally called division-of-focal-plane (DoFP) cameras, include the Bayer color filter array (CFA) design common to commercial three-color cameras. Spatial filtering is accomplished by repeating a spatial filter pattern across the focal plane array. In CFA cameras, each detector element only collects a single color channel, resulting in a mosaic pattern. The other color channels must be estimated at that sample location. This is accomplished by a process called, demosaicing. Extensive research has been accomplished in demosaicing. However, these processes are typically insufficient in eliminating aliasing artifacts. The research described in this work is developed to overcome sampling limitations in DoFP cameras and reduce or eliminate aliasing in images. Two new methods are developed and are presented. Both methods create SR image estimates from multiple image frames in the presence of random frame-to-frame motion. The first method is an interpolation-restoration SR approach that places all frame samples onto a common non-uniform sampling space and applies an optimal filter based on sample pattern and image statistics. This fast method combines multiple SR steps as a spatially-adaptive weighted sum and is called the color adaptive Wiener filter (AWF) SR approach. By leveraging a newly-developed approach to capture channel cross-correlation, this fast approach rivals and exceeds performance of much more computationally-intensive variational methods. The second method in this work is a new and c (open full item for complete abstract)

Master of Science, The Ohio State University, 2010, Geodetic Science and Surveying

Image retrieval has applications in different disciplines. For example, there are applications in digital painting catalogues and in security related applications Researchers from both computer vision and photogrammetry fields are developing robust image retrieval methods that can be used for achieving, browsing and searching. Various approaches have been developed by researchers to solve the retrieval problem using different image features, including color, texture and shape of objects in the image. Our method is motivated from a geometric invariance framework, which is based on invariance of conic sections under the projective image transformation. First, conic sections, which are fitted to object boundaries, are generated. The invariance property of these conic sections is used to represent the shape of the object boundaries. This representation provides an invariant signature of that image. Once an invariant signature is obtained for each image, certain classification methods are used to test whether these signatures present unique characteristics for each image group. Additionally, a retrieval mechanism is built that uses invariant signatures of each image to build a relationship with other images and to retrieve the most related ones. A measure of the relationship between images is obtained by using two common metrics histogram intersection and minimum pair distance assignment. It is hypothesized in this research that generated invariant signatures present unique characteristics for each image group and these signatures can be used for classification and retrieval of images in a database. This hypothesis is satisfied in terms of classification, but it is not satisfied for retrieval problems because of degenerate conics.

Doctor of Philosophy, The Ohio State University, 2010, Electrical and Computer Engineering

The recent developments in whole-slide digital scanners have spurred a revolution in imaging technology for histopathology. While these commercially available, high-throughput whole-slide scanners address data acquisition issues, the amount of data provided by them currently far exceeds the rate at which they can be analyzed efficiently. More importantly, the qualitative microscopic visual inspection of tissue slides by human readers (e.g., pathologists) is often subject to significant inter- and intra-reader variations. Using computerized image analysis, it is possible to extract more objective and precise quantitative diagnostic clues that will help improving the current evaluation of histopathological data. The main goal of this dissertation is to understand and address the challenges associated with the development of image analysis techniques for the computer-aided interpretation of high-resolution histopathology imagery. We aim to design algorithms for key image analysis tasks such as robust and adaptive segmentation of cytological components for higher level processing, construction of biologically relevant and computationally tractable features and their mathematical representations in order to differentiate distinct tissue subtypes, detection of prognostically significant tissue structures, and spatial alignment of tissue sections prepared with different stains in order to incorporate complementary information. We demonstrate the effectiveness of the proposed approaches on three important histopathology applications: analysis of whole-slide tissue sections for neuroblastoma prognosis, automated grading of follicular lymphoma and quantitative characterization of muscle fiber subtypes from serial transverse skeletal muscle tissue samples. For computer-aided analysis of whole-slide neuroblastoma tissue sections, we develop a comprehensive, multi-resolution image analysis framework including the establishment of multi-resolution image hierarchy, image segmentat (open full item for complete abstract)

Doctor of Philosophy, The Ohio State University, 2007, Electrical Engineering

Because of the contradiction of the vast data size of raw digital images and videos and the limited transmission bandwidth and storage space, it is essential to develop compression methodologies with high compression ratio and good reconstructed quality. It is also important to develop quality metrics which are consistent with human vision and easy to calculate. The spatial-frequency localization and multi-resolution capabilities of the wavelet transform make it a natural means of signal representation. This work investigates the advantages of the wavelet transform and focuses on the following research topics: 1) An image quality metric that assesses the quality of an image in the wavelet domain; 2) A quality constrained compression algorithm that compresses an image to a desired visual quality; 3) An innovative DWT-based temporal filtering scheme that achieves high compression ratio and reduces the ghost effect without motion estimation; 4) A virtual sub-object video coding scheme that is suitable for applications with static background.