Sparse visual signal representations and selected applications

Sparse representation has been well investigated and discussed over the past decade due to its ability in visual signal discrimination for various applications such as face recognition, image classification and video clustering. It has attracted more and more interest in the recent years because of...

Full description

Saved in:
Bibliographic Details
Main Author: Hung, Tzu-Yi
Other Authors: Tan Yap Peng
Format: Theses and Dissertations
Language:English
Published: 2015
Subjects:
Online Access:https://hdl.handle.net/10356/65048
Tags: Add Tag
No Tags, Be the first to tag this record!
Institution: Nanyang Technological University
Language: English
Description
Summary:Sparse representation has been well investigated and discussed over the past decade due to its ability in visual signal discrimination for various applications such as face recognition, image classification and video clustering. It has attracted more and more interest in the recent years because of the increasing demands for developing real world systems with large-scale image and video collections. While a large number of sparse representation algorithms have been proposed in the literature and some encouraging results have been obtained, there is still a need for further improvement. This thesis aims to address various issues of sparse representation, including feature quantization models, sparsity estimation methods and dictionary learning techniques for sparse visual signal representation over different computer vision and pattern recognition tasks such as image classification, action recognition and activity-based human identification to demonstrate their efficacy and superiority over state-of-the-art methods. More specifically, we focus our work on two directions: 1) An application-oriented problem: we investigate the problem of activity-based person identification which will be elaborated in the thesis; and 2) A model-oriented problem: we improve the existing sparse coding approaches in a more efficient and effective way and evaluate the performance of the proposed method on several visual tasks.