Object localization is an important task in computer vision, which is usually handled by searching for an optimal subwindow that tightly covers the object of interest. Both boundary-based shape and region-based appearance features are important to accurate object localization. For some objects, shape feature might be more important and for some objects, appearance feature might be more important. However, current state-of-the-art object localization methods either focus on shape feature or focus on appearance feature, and efficiently combining shape and appearance features to achieve object localization is a very challenging research topic in computer vision. In addition, the subwindows considered in previous work are usually limited to rectangles or other specified, simple shapes. With such specified shapes, there may not exist a subwindow that can cover the object of interest tightly. As a result, the desired subwindow around the object of interest may not be optimal in terms of the localization objective function and cannot be detected by the subwindow search algorithm. In my dissertation, to address the above problems we proposed new approaches to combine shape and appearance features for object localization, in a globally optimal fashion, using graph-theoretic models and algorithms. We first develop an edge grouping based free-shape subwindow search algorithm for object localization, where no specific shape features of individual object classes are considered. We just generally require the bounding contour (free-shape subwindow) to be aligned with detected edges and cover the desired object appearance features that are learned from a training set. This requirement is quantified and integrated into the localization objective function based on the widely-used bag of visual words technique. We then extend the edge grouping based free shape subwindow search method to super-edge grouping method, where both the shape and appearance features of specific object classes are learned and then integrated to the object localization algorithm. Experiments show that our proposed method, by integrating both boundary-based shape feature and region-based appearance feature, can produce better localization performance than the previous state-of-the-art subwindow search methods.
","abstract_html":"<p>Object localization is an important task in computer vision, which is usually handled by searching for an optimal subwindow that tightly covers the object of interest. Both boundary-based shape and region-based appearance features are important to accurate object localization. For some objects, shape feature might be more important and for some objects, appearance feature might be more important. However, current state-of-the-art object localization methods either focus on shape feature or focus on appearance feature, and efficiently combining shape and appearance features to achieve object localization is a very challenging research topic in computer vision. In addition, the subwindows considered in previous work are usually limited to rectangles or other specified, simple shapes. With such specified shapes, there may not exist a subwindow that can cover the object of interest tightly. As a result, the desired subwindow around the object of interest may not be optimal in terms of the localization objective function and cannot be detected by the subwindow search algorithm. In my dissertation, to address the above problems we proposed new approaches to combine shape and appearance features for object localization, in a globally optimal fashion, using graph-theoretic models and algorithms. We first develop an edge grouping based free-shape subwindow search algorithm for object localization, where no specific shape features of individual object classes are considered. We just generally require the bounding contour (free-shape subwindow) to be aligned with detected edges and cover the desired object appearance features that are learned from a training set. This requirement is quantified and integrated into the localization objective function based on the widely-used bag of visual words technique. We then extend the edge grouping based free shape subwindow search method to super-edge grouping method, where both the shape and appearance features of specific object classes are learned and then integrated to the object localization algorithm. Experiments show that our proposed method, by integrating both boundary-based shape feature and region-based appearance feature, can produce better localization performance than the previous state-of-the-art subwindow search methods.</p>","abstract_has_math":false,"creators":["Zhang, Zhiqi"],"institution":null,"degree_name":"Ph.D.","degree_level":"Campus Access Dissertation","degree_discipline":"Computer Science and Engineering","degree_department":null,"school":null,"contributors":["Wang, Song"],"advisors":[],"committee_chairs":[],"committee_members":[],"year":2012,"date_issued":"2012-01-01T08:00:00Z","date_published":"2012-01-01T08:00:00Z","updated_at":"2026-07-24T04:38:14Z","subjects":["Computer Sciences","Electrical and Computer Engineering","Engineering","Physical Sciences and Mathematics","Computer Vision","Edge Grouping","Image segmentation","Object detection","Object Localization","Super Edge"],"languages":[],"rights":["© 2012, Zhiqi Zhang"],"rights_urls":[],"identifier_entries":[]},"links":{"outbound_url":"https://scholarcommons.sc.edu/etd/806","outbound_label":"Repository record","outbound_source":"dc:identifier"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor","label":"Contributor","values":["Wang, Song"]},{"key":"dc:creator","label":"Author","values":["Zhang, Zhiqi"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"thesis:degree_discipline","label":"Discipline","values":["Computer Science and Engineering"]},{"key":"thesis:degree_level","label":"Degree Level","values":["Campus Access Dissertation"]},{"key":"thesis:degree_name","label":"Degree Name","values":["Ph.D."]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Computer Sciences","Electrical and Computer Engineering","Engineering","Physical Sciences and Mathematics","Computer Vision","Edge Grouping","Image segmentation","Object detection","Object Localization","Super Edge"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:rights","label":"Dc Rights","values":["© 2012, Zhiqi Zhang"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier","label":"Identifier","values":["https://scholarcommons.sc.edu/etd/806"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["Object localization is an important task in computer vision, which is usually handled by searching for an optimal subwindow that tightly covers the object of interest. Both boundary-based shape and region-based appearance features are important to accurate object localization. For some objects, shape feature might be more important and for some objects, appearance feature might be more important. However, current state-of-the-art object localization methods either focus on shape feature or focus on appearance feature, and efficiently combining shape and appearance features to achieve object localization is a very challenging research topic in computer vision. In addition, the subwindows considered in previous work are usually limited to rectangles or other specified, simple shapes. With such specified shapes, there may not exist a subwindow that can cover the object of interest tightly. As a result, the desired subwindow around the object of interest may not be optimal in terms of the localization objective function and cannot be detected by the subwindow search algorithm. In my dissertation, to address the above problems we proposed new approaches to combine shape and appearance features for object localization, in a globally optimal fashion, using graph-theoretic models and algorithms. We first develop an edge grouping based free-shape subwindow search algorithm for object localization, where no specific shape features of individual object classes are considered. We just generally require the bounding contour (free-shape subwindow) to be aligned with detected edges and cover the desired object appearance features that are learned from a training set. This requirement is quantified and integrated into the localization objective function based on the widely-used bag of visual words technique. We then extend the edge grouping based free shape subwindow search method to super-edge grouping method, where both the shape and appearance features of specific object classes are learned and then integrated to the object localization algorithm. Experiments show that our proposed method, by integrating both boundary-based shape feature and region-based appearance feature, can produce better localization performance than the previous state-of-the-art subwindow search methods.
"]},{"key":"dc:title","label":"Title","values":["Object Localization by Combining Shape and Appearance Features"]}]}],"canonical_facts":{"dc:contributor":["Wang, Song"],"dc:creator":["Zhang, Zhiqi"],"dc:description.abstract":["Object localization is an important task in computer vision, which is usually handled by searching for an optimal subwindow that tightly covers the object of interest. Both boundary-based shape and region-based appearance features are important to accurate object localization. For some objects, shape feature might be more important and for some objects, appearance feature might be more important. However, current state-of-the-art object localization methods either focus on shape feature or focus on appearance feature, and efficiently combining shape and appearance features to achieve object localization is a very challenging research topic in computer vision. In addition, the subwindows considered in previous work are usually limited to rectangles or other specified, simple shapes. With such specified shapes, there may not exist a subwindow that can cover the object of interest tightly. As a result, the desired subwindow around the object of interest may not be optimal in terms of the localization objective function and cannot be detected by the subwindow search algorithm. In my dissertation, to address the above problems we proposed new approaches to combine shape and appearance features for object localization, in a globally optimal fashion, using graph-theoretic models and algorithms. We first develop an edge grouping based free-shape subwindow search algorithm for object localization, where no specific shape features of individual object classes are considered. We just generally require the bounding contour (free-shape subwindow) to be aligned with detected edges and cover the desired object appearance features that are learned from a training set. This requirement is quantified and integrated into the localization objective function based on the widely-used bag of visual words technique. We then extend the edge grouping based free shape subwindow search method to super-edge grouping method, where both the shape and appearance features of specific object classes are learned and then integrated to the object localization algorithm. Experiments show that our proposed method, by integrating both boundary-based shape feature and region-based appearance feature, can produce better localization performance than the previous state-of-the-art subwindow search methods.
"],"dc:identifier":["https://scholarcommons.sc.edu/etd/806"],"dc:rights":["© 2012, Zhiqi Zhang"],"dc:subject":["Computer Sciences","Electrical and Computer Engineering","Engineering","Physical Sciences and Mathematics","Computer Vision","Edge Grouping","Image segmentation","Object detection","Object Localization","Super Edge"],"dc:title":["Object Localization by Combining Shape and Appearance Features"],"thesis:degree_discipline":["Computer Science and Engineering"],"thesis:degree_level":["Campus Access Dissertation"],"thesis:degree_name":["Ph.D."]},"updated_at":"2026-07-24T04:38:14Z"}