Video sharpness based barcodes with Bessel functions and quaternions for video summarization

Loading...
Thumbnail Image
Date
2022-03-28
Authors
Pilippu Hewa Don, Arjuna
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Pertaining to the increasing amount of video data created every day, fast and effective methods of analyzing videos with minimum human supervision is a requirement in the current world. Several methods have been introduced by the research community to tackle this problem, but they have shown certain shortcomings when it comes to real-life deployment. The main objective of this research is to develop a set of tools for recognizing and isolating parts of a video that are important, which will reduce the amount of time required for the analysis. One way of understanding the important parts of a video is to isolate frames that have important information. The other way is to find the area of the video frame that has the most important information. To solve these two problems, this thesis considers segmentation of video frame foreground from the background using image sharpness. This research also introduces improved Ghrist barcodes as a method of presenting this information. A Ghrist barcode is a topology of data pictograph useful in representing the persistence of the features of changing shapes. These barcodes show the timeline information where there is an activity of interest. Bessel functions and quaternions are suggested as a method to find the area of interest inside a video frame using the change in sharpness of the video frames. Bessel functions are more flexible in measuring sharpness allowing various applications with minimum modifications. With quaternions, it is possible to obtain three dimensional (RGB) perspective for color distance which is an update over the traditional grayscale-based object sharpness. These methods have been used to introduce updated object sharpness methodologies that are both robust and fast. Object sharpness paves the way for topological analysis of the object shapes and their changes inside the video frame over time. The suggested methods have been applied in real-life applications and the obtained results are promising.
Description
Keywords
Image Sharpness, Video Summarization, Video Barcodes, Bessel Functions, Quaternions
Citation