以二維隱藏式馬可夫模型為基礎的基礎的影像整體檢索

Viual Apprearance Retrieval Based on the 2-dimensional Pseudo Hidden Markov Model

張敏達(Min-Ta Chang)

指導教授 ： 陳淑媛

元智大學/工程學院/電機與資訊工程研究所/碩士(1997年)

摘要

本論文擬提出一個以虛擬二維隱藏式馬可夫模型為基礎的影像整體外觀檢索法。
以內容為基礎的影像檢索法大多著眼於色彩特徵或形狀輪廓，但是以輪廓為基準
的檢索方法首先(1)需要克服相似影像但輪廓不同時的問題；而以色彩為主的檢索
方法缺乏對輪廓的檢索能力。本文希望能結合上述兩者的優點，唯形狀特徵著眼
整體外觀以避開比對影像中組成區塊的形狀或個數不同的問題。另一方面，由於
--本法採用(2)具有高度容錯能力的二維隱藏式馬可夫模型，是故比對結果非常具
有彈性。此外，(3)本法亦採用對數極座標取樣，以獲得旋轉，大小比例變化之不
變性。最後，本論文還提出一套簡單的彩色滅減法以降低彩色處理時的運算量。
本文方法以驗證過商標資料庫，且實驗結果證實本法確實可行。

關鍵字

二維隱藏式馬可夫模型 ； 對數極座標對映 ； 顏色量化 ； 影像檢索 ； 商標辨識

並列摘要

A new appearance retrieval method based on the two-dimensional pseudo hidden Markov model (2D PHMM) is proposed.
Most content-based retrieval system focus on colour features or shape silhouettes. However, it is hard to use silhouette-based
approaches to retrieve images which have a variety of numbers of decomposed components. Thus, our method focus on the
whole appearance, including colour and shape features but not silhouettes, to tackle the differences in the numbers of components
included in various images. On the other hand, 2-diminsional pseudo hidden Markov model (2D PHMM) has high tolerance
to noise and distortion, thus, the flexibility of pattern matching can be achieved by the adopted 2D PHMM. In addition, the
log-polar mapping is adopted in this study to overcome rotation and scaling invariance requirement. Moreover, a simple colour
quantization method is proposed to reduce the tremendous computation load as required by colour processing. Experiments has
been conducted on a trademark database to prove the feasibility of the proposed method.

並列關鍵字

2D PHMM ； log-polar mapping ； color quantization ； image retrieval ； trademark recognition

參考文獻

[1]M. J. Swain and D. H. Ballard, "Color indexing," Int. J. Computer Vision, vol. 7, pp. 11-32, 1991.

[2]B. M. Mehtre et al., "Color matching for image retrieval," Pattern Recognition Letters, vol. 16, pp. 325-331, 1995.

[3]M. S. Kankanhali et al., "Cluster-based color matching for image retrieval," Pattern Recognition, vol. 29, no. 4, pp. 701-708, 1996.

[4]A. K. Jain and A. Vailaya, "Image retrieval using color and shape," Pattern Recognition, vol. 29, no. 8, pp. 1233-1244, 1996.

[5]H. C. Lin et al., "Color image retrieval based hidden Markov models," IEEE Trans. on Image Processing, vol. 6, no. 2, 1997.