使用作用力分佈圖於乏晰空間關係之檢索

Using Force Histogram in Retrieving Fuzzy Spatial Relationship

粘智超(Chi-chiao Nien)

指導教授 ： 周清江

淡江大學/管理學院/資訊管理學系碩士班/碩士(2005年)

https://doi.org/10.6846/TKU.2005.00352

摘要

隨著數位影像處理工具的日益普及，產生了大量豊富且多樣的影像資料，如何有效的從龐大的影像資料庫中檢索出使用者所要的影像，成為影像檢索上重要的課題。早期的影像檢索是透過影像的文字說明，但影像資料的描述，若用傳統人工輸入方式建置，不但較為主觀，而且需花費太多的時間、金錢和人力去管理，因此陸續有學者提出以影像本身的內容，如顏色、紋理、物件之形狀、物件之空間關係等等，作為影像檢索之基礎。
本論文針對影像內容中物件之空間關係，提出以物件間作用力分佈圖為基礎之數個乏晰空間關係特徵值，並將這些值作為相似比對之依據。這種方法與過去直接利用分佈圖配對的方式相較，優點在於(1)計算效率上較快速;(2)利用擷取出之空間關係特徵值建立索引，可在檢索相似配對時，節省與資料庫影像比對的時間;(3)這些特徵值具有語意上認知之合理性。本論文以實驗說明我們所建置之方向、包圍、遠近等乏晰空間關係，能較完整展現影像物件間空間關係之差異性，未來應可用於語意式影像檢索系統中。

關鍵字

作用力分佈圖 ； 空間關係 ； 乏晰理論

並列摘要

With the popularity of digital image generation and processing tools, huge miscellaneous rich image data have been produced. How to effectively retrieve images from the huge image databases has become an important subject. In the early stage, image retrieval was achieved by matching keywords with image description text. However, the manual input of image description is not only too subjective, but also spends a lot of time, money and manpower. Thus, several researchers proposed successively retrieval methods based on the image content, such as color, texture, shapes of objects, spatial relationships of objects, etc.

To obtain a better matching of spatial relationships, we propose several fuzzy spatial relationship characteristic values based on the force histograms among the objects in the images. These values are further used to compute the similarity of two images. This method, compared with direct histogram matching, has the following advantages: (1) It has better computational efficiency; (2) It could precompute the characteristic values of the spatial relationships and and store them in the database, which tremendously saves time in retrieving similar images; (3) These characteristic values are associated with more human-reasonable semantic meanings. Lastly, we demonstrate the use of fuzzy directional, surrounding and distance spatial relationships in image retrieval. The results illustrate that these fuzzy spatial relationships can extract the difference of the spatial relationship among the images more completely. We hope this system could be applied to semantic retrieval of the images in the future.

並列關鍵字

Force-Histogram ； Spatial Relationship ； Fuzzy Theory

參考文獻

1. Bloch, I., Ralescu, A., “Directional Relative Position Between Objects in Image Processing: A Comparison Between Fuzzy Approaches,” Pattern Recognition, 2003, Vol. 36, No. 7, pp. 1563-1582.

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

3. Chang, C. C., “Spatial Match Retrieval of Symbolic Pictures,” Journal of Information Science and Engineering, 1991, Vol. 7, pp. 405-422.

4. Gader, P. D., “Fuzzy Spatial Relations Based on Fuzzy Morphology,” Proceedings of the 6th IEEE International Conference on Fuzzy Systems, 1997, Vol. 2, pp. 1179-1183.

5. Gudivada, N., Raghavan, V., “Design and Evaluation of Algorithms for Image Retrieval by Spatial Similarity,” ACM Transactions on Information Systems, 1995, Vol. 13, No. 2, pp. 115-144.