Similarity Retrieval by Using Spatio-temporal Relationships in Video databases
碩士 === 國立臺灣大學 === 資訊管理研究所 === 91 === Usually a video consists of a number of frames and each frame may contain several objects. The changes of video objects in spatial and temporal dimension are quite useful to measure two videos. In this thesis, we propose a new method to measure the sim...
| Main Authors: | , |
|---|---|
| Other Authors: | |
| Format: | Others |
| Language: | en_US |
| Published: |
2003
|
| Online Access: | http://ndltd.ncl.edu.tw/handle/51621255511952398769 |
| Summary: | 碩士 === 國立臺灣大學 === 資訊管理研究所 === 91 === Usually a video consists of a number of frames and each frame may contain several objects. The changes of video objects in spatial and temporal dimension are quite useful to measure two videos. In this thesis, we propose a new method to measure the similarity (dissimilarity) between two videos by spatio-temporal relationships between objects. The new method basically includes two phases. One is to split a video into several intervals of video and make use of the finding maximal similar object-set algorithm to compute the maximal similar object-sets as the similarity norm between any two intervals. The other is to utilize the maximal similar object-sets between intervals obtained from the first phase to compute maximal similar object-sets between videos in the dynamic programming approach. We propose continuous and discontinuous intervals matching algorithm for this part. Some experiments are performed to show the efficiency and effectiveness of our proposed algorithm.
|
|---|