Sponsor
Portland State University. Department of Computer Science
First Advisor
Wu-chi Feng
Term of Graduation
Fall 2024
Date of Publication
12-16-2024
Document Type
Dissertation
Degree Name
Doctor of Philosophy (Ph.D.) in Computer Science
Department
Computer Science
Language
English
Subjects
Adaptive Image Acquisition, Efficient Visual Data Processing, Hybrid Video Compression, Learning-based Compression, Resource Utilization Improvement, Single-Photon Cameras (SPCs)
Physical Description
1 online resource (xiii, 144 pages)
Abstract
With continuous advancements in technology, the volume of visual data being captured, distributed, and consumed across various applications is rapidly increasing. Enhancing the efficiency of visual data processing can improve resource consumption, making the storage, transmission, and use of this growing visual content more effective.
Image and video data constitute a significant share of internet traffic, making video compression a critical factor in enhancing overall data throughput by improving the video compression ratio. Capturing, transferring, and storing raw video data is challenging due to the substantial resources required for both storage and computation. Video compression, however, can significantly mitigate these demands. To achieve this, it is essential to employ an efficient compression approach that maintains high visual quality frames while reducing transmission rates. In this dissertation, we propose two video compression frameworks designed to reduce bandwidth consumption while preserving, and in some cases even improving, visual quality compared to other approaches.
Moreover, efficient visual data acquisition and processing is essential for conserving resources in camera systems. This visual data can be captured across a range of systems, encompassing both conventional and unconventional cameras. Conventional cameras typically rely on CCD or CMOS sensors, while unconventional cameras may use alternatives such as single-photon cameras (SPCs). In some challenging scenarios--such as high-dynamic range imaging, ultra-high-speed photography, and low-light conditions--these conventional sensors fall short of the performance capabilities offered by SPCs. Most visual data generated today relies on conventional sensors, such as CCD and CMOS, supported by efficient processing and compression techniques to manage the data they produce.
In contrast, data captured by SPCs differs in format from that of conventional sensors, necessitating the development of efficient methods for capturing, processing, and compressing single-photon camera data. Unlike pixels in conventional sensors, which provide a single 8 to 16-bit brightness value, SPCs can timestamp millions of individual photon arrivals, leading to bottlenecks in their pixel circuit. In this dissertation, we design and develop an image acquisition model to simulate SPC behavior. Additionally, we propose adaptive algorithms that efficiently allocate hardware resources across groups of pixels to mitigate the bottlenecks by selectively subsampling data entering the sensor. This approach reduces the load on the underlying systems while preserving the visual quality of the resulting visual data.
Rights
© 2024 Yeganeh Jalalpour
In Copyright. URI: http://rightsstatements.org/vocab/InC/1.0/ This Item is protected by copyright and/or related rights. You are free to use this Item in any way that is permitted by the copyright and related rights legislation that applies to your use. For other uses you need to obtain permission from the rights-holder(s).
Persistent Identifier
https://archives.pdx.edu/ds/psu/42897
Recommended Citation
Jalalpour, Yeganeh, "Efficient Visual Data Processing Approaches to Improve Resource Utilization" (2024). Dissertations and Theses. Paper 6735.