根據物件間空間關係變化之影片相似尋取

Similarity retrieval of videos based on the change of the spatial relationship between objects

陳顗任(Yi-Ren Chen)

指導教授 ： 梁恩輝

淡江大學/商管學院/資訊管理學系碩士班/碩士(2014年)

https://doi.org/10.6846/TKU.2014.00264

摘要

一個影片的內容可以視為一連串的影格，而每一個影格可以經由轉換以一個符號影像(symbolic image)來表示。根據符號影像中物件間空間關係之影片相似尋取是一個影片尋取的重要方法。在先前許多影片相似尋取的研究中，大多是根據影像中物件之空間關係計算查詢影片中之影像與參考影片中之影像的相似度，並以此作為計算查詢影片與參考影片相似程度之基礎。然而在影片中通常物件會移動，因此物件間之空間關係會產生變化，此種方法並未直接明確的考慮到在影片中物件間之空間關係的變化。
在本論文中，我們將先定義物件間九種空間關係變化的類型。在影片中，從一影格至其下一影格，每兩個物件之間就會發生一種類型的空間關係變化，當多個物件同時存在時，在相鄰兩影格間就會有一組空間關係的變化，稱為空間關係變化集合(the Integrated Spatial Relationship Change，簡稱ISRC)。我們提出計算兩ISRC間之相似度方法。影片的相似尋取可以根據一串查詢的ISRC進行，將查詢ISRC序列和參考影片之ISRC序列做比較，並根據ISRC間之相似度，我們提出在參考影片之ISRC序列中找出與查詢影片相似的片段的方法。如此一來，即可進行以物件間之空間關係的變化為基礎之查詢。

關鍵字

符號影像 ； 影片相似尋取 ； 空間關係的變化

並列摘要

A video can be viewed as a sequence of frames. Each frame can be transformed into a symbolic image. Similarity retrieval of video based on the spatial relationship between objects in the symbolic image is an important method for video retrieval. In the previous research, the similarity between the images of the query video and that of the reference video according to the spatial relationship between objects is computed, and then it is used as the base for computing the similarity between the query video and the reference video. However, objects usually move in the video and the spatial relationships between objects may change. The change of the spatial relationship between objects in the video is not considered directly and precisely in previous research.
In this thesis, we define nine types of changes of spatial relationship between objects. In the video, one type of change between two objects occurs from one frame to the next. When multiple objects exist, there will be a set of spatial relations between the two adjacent frames, named as the integrated spatial relationship change (ISRC). We propose a method to calculate the similarity between two ISRCs. Similarity retrieval of video can be performed according to a sequence of query ISRC. The query ISRC sequence is compared with the ISRC sequence of the reference video. We proposed a method to locate the segment in the reference video similar to the query ISRC sequence based on the similarity between ISRCs. Hence, Similarity retrieval of video based on the change of spatial relationship can be achieved.

並列關鍵字

Symbols Images ； Similarity Retrieval ； Change of Spatial Relationship

參考文獻

[3] Chang, S.K., Jungert, E., and Li Y., “Representation and retrieval of symbolic pictures using generalized 2D strings,” in: SPIE Proceedings on Visual Communications and Image Processing, Philadelphia, pp. 1360-1372, 1989.

[4] Chang, S.K., Shi, Q.Y., and Yan, C.W., “Iconic indexing by 2-D strings,” IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. PAMI-9, No. 3, pp. 413-428, 1987.

[7] Huang, P.W., and Jean, Y.R., “Using 2D C+-string as spatial knowledge representation for image database systems,” Pattern Recognition, Vol. 27, No. 9, pp. 1249–1257, 1994.

[8] Jungert, E., “Extended symbolic projections as a knowledge structure for spatial reasoning,” in: Proceedings of 4th BPRA Conference on Pattern Recognition, pp. 343–351, 1988.

[9] Lee, A.J.T., and Chiu, H.P., “2D Z-string: a new spatial knowledge representation for image databases,” Pattern Recognition Letters, Vol. 24, No. 16, pp. 3015–3026, 2003.