Shape retrieval and matching based on line segments.
A novel local structure based image retrieval (ALSBIR) approach is proposed in this thesis to build a general framework for object / image retrieval. It performs speedy interpretation of input images and retrieval of structurally similar models from large database according to the input. It is first...
Saved in:
Main Author: | |
---|---|
Other Authors: | |
Format: | Theses and Dissertations |
Published: |
2008
|
Subjects: | |
Online Access: | https://hdl.handle.net/10356/2478 |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Institution: | Nanyang Technological University |
Summary: | A novel local structure based image retrieval (ALSBIR) approach is proposed in this thesis to build a general framework for object / image retrieval. It performs speedy interpretation of input images and retrieval of structurally similar models from large database according to the input. It is first of its kind to experiment with mixed-object query in shape retrieval. The approach consists of a novel local shape representation inspired by Gestalt psychology and a novel hypercube indexing structure driven by dynamic programming. It can tackle partially occluded objects in a cluttered environment since it employs local information only and does not require separation of whole input objects from complex background. It is able to infer retrieval results from incomplete information of an input by first extracting consistent and structurally unique local neighborhood information from inputs or models, and then voting on the optimal matches. The proposed concepts have been compared with the 6 nearest-neighbors shape description. They have the similar performance on single object retrieval. However, the proposed method out-performed the 6 nearest-neighbors on the test of occluded and cluttered object queries. The sensitivity analysis and error analysis shows that the system is robust in the occluded and cluttered environments. |
---|