University of British Columbia
Hierarchical segment boost : a segment level classification approach to object class recognition
Abstract
dc:descriptionIn this work we address the problem of object recognition and localization within cluttered, natural scenes. Specifically we present a new approach to recognizing deformable objects that uses only local information. We suggest a new method of computing labels for arbitrary regions within an image using only local color and texture information. The results demonstrate our success in both identifying and localizing several classes of objects within cluttered scenes. We make two primary contributions to the field of deformable object recognition. First we present a new technique for labeling arbitrary regions within an image using texture and color features. Second we introduce a hierarchical approach to combining the classification results from segmentations of different granularity. Since the field of deformable object recognition is still in its infancy, we hope the work presented here will be used as a stepping stone to developing more complex, multi-object detection systems.
Degree
thesis:*- Name thesis:degree_name
- Master of Science - MSc
- Level thesis:degree_level
- master's
- Discipline thesis:degree_discipline
- Computer Science
- Grantor dc:publisher
- University of British Columbia
- Year dc:date
- 2006
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Reynolds, Jordan Michael
Rights
dc:rights- Statement dc:rights
-
- For non-commercial purposes only, such as research, private study and education. Additional conditions apply, see Terms of Use https://open.library.ubc.ca/terms_of_use.
- Language dc:language
- eng
Identifiers
dc:identifier.*- Handle dc:identifier
- http://hdl.handle.net/2429/32286