Optimized event-based filtering algorithms for hardware implementation
With the development of modern photography, scientists have invented many kinds of cameras that use a large range of techniques, like the Single Lens Reflex camera, electron microscope and high-speed cameras. These cameras capture images at specific moments which means they record the absolute value...
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Thesis-Master by Coursework |
| Language: | English |
| Published: |
Nanyang Technological University
2022
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/159300 |
| _version_ | 1826114843596292096 |
|---|---|
| author | Luo, Zhiwen |
| author2 | Chang Chip Hong |
| author_facet | Chang Chip Hong Luo, Zhiwen |
| author_sort | Luo, Zhiwen |
| collection | NTU |
| description | With the development of modern photography, scientists have invented many kinds of cameras that use a large range of techniques, like the Single Lens Reflex camera, electron microscope and high-speed cameras. These cameras capture images at specific moments which means they record the absolute value of light at those moments. The phenomenon of ”persistence of vision” makes it possible to form a video using a series of photographs shot by these cameras. However, such a technique brings with it the serious problem that the limited frame rate can result in videos with motion blur. Specialized high-speed cameras have been designed to solve this problem, and can produce up to thousands or millions frames per second. Though high-speed cameras exhibit better performance, they are much more expensive and less portable compared to other cameras. All the cameras mentioned above can be classified as “frame-based” cameras as they capture the image “frames” at a steady rate and record the absolute value of light intensity at each pixel.
A new type of camera has been designed to find a fundamental solution to the motion blur issue. The new camera imitates the function of biological vision systems, sensing the change of light intensity at each pixel of the sensors. Light intensity change in each pixel is independent of the changes in other pixels and each change in light intensity generates a new “event” from the camera. The camera is therefore called an “event-based” camera.
The event-based camera is sensitive to changes in light intensity and fundamentally avoids motion blur because the output is not a complete image but a series of events containing signal and noise. As in the other cameras, denoising important to event-based cameras because it is the very first step of whole image processing chain.
In this dissertation, several denoising techniques for event-based cameras are introduced, and some of them like the Background Activity Filter and O(N) filter are discussed in detail. An improved version of the BAF filter is also introduced and implemented in FPGA. |
| first_indexed | 2024-10-01T03:45:49Z |
| format | Thesis-Master by Coursework |
| id | ntu-10356/159300 |
| institution | Nanyang Technological University |
| language | English |
| last_indexed | 2024-10-01T03:45:49Z |
| publishDate | 2022 |
| publisher | Nanyang Technological University |
| record_format | dspace |
| spelling | ntu-10356/1593002023-07-04T17:53:07Z Optimized event-based filtering algorithms for hardware implementation Luo, Zhiwen Chang Chip Hong School of Electrical and Electronic Engineering ECHChang@ntu.edu.sg Engineering::Electrical and electronic engineering With the development of modern photography, scientists have invented many kinds of cameras that use a large range of techniques, like the Single Lens Reflex camera, electron microscope and high-speed cameras. These cameras capture images at specific moments which means they record the absolute value of light at those moments. The phenomenon of ”persistence of vision” makes it possible to form a video using a series of photographs shot by these cameras. However, such a technique brings with it the serious problem that the limited frame rate can result in videos with motion blur. Specialized high-speed cameras have been designed to solve this problem, and can produce up to thousands or millions frames per second. Though high-speed cameras exhibit better performance, they are much more expensive and less portable compared to other cameras. All the cameras mentioned above can be classified as “frame-based” cameras as they capture the image “frames” at a steady rate and record the absolute value of light intensity at each pixel. A new type of camera has been designed to find a fundamental solution to the motion blur issue. The new camera imitates the function of biological vision systems, sensing the change of light intensity at each pixel of the sensors. Light intensity change in each pixel is independent of the changes in other pixels and each change in light intensity generates a new “event” from the camera. The camera is therefore called an “event-based” camera. The event-based camera is sensitive to changes in light intensity and fundamentally avoids motion blur because the output is not a complete image but a series of events containing signal and noise. As in the other cameras, denoising important to event-based cameras because it is the very first step of whole image processing chain. In this dissertation, several denoising techniques for event-based cameras are introduced, and some of them like the Background Activity Filter and O(N) filter are discussed in detail. An improved version of the BAF filter is also introduced and implemented in FPGA. Master of Science (Electronics) 2022-06-14T06:05:31Z 2022-06-14T06:05:31Z 2022 Thesis-Master by Coursework Luo, Z. (2022). Optimized event-based filtering algorithms for hardware implementation. Master's thesis, Nanyang Technological University, Singapore. https://hdl.handle.net/10356/159300 https://hdl.handle.net/10356/159300 en application/pdf Nanyang Technological University |
| spellingShingle | Engineering::Electrical and electronic engineering Luo, Zhiwen Optimized event-based filtering algorithms for hardware implementation |
| title | Optimized event-based filtering algorithms for hardware implementation |
| title_full | Optimized event-based filtering algorithms for hardware implementation |
| title_fullStr | Optimized event-based filtering algorithms for hardware implementation |
| title_full_unstemmed | Optimized event-based filtering algorithms for hardware implementation |
| title_short | Optimized event-based filtering algorithms for hardware implementation |
| title_sort | optimized event based filtering algorithms for hardware implementation |
| topic | Engineering::Electrical and electronic engineering |
| url | https://hdl.handle.net/10356/159300 |
| work_keys_str_mv | AT luozhiwen optimizedeventbasedfilteringalgorithmsforhardwareimplementation |