利用仿射不變特徵改進穩定極值區域之比對效能

Matching Performance Improvement of Maximally Stable Extremal Region with Local Affine Invariants

劉祐寧(Yu-Ning Liu)

指導教授 ： 貝蘇章

國立臺灣大學/電機資訊學院/電信工程學研究所/碩士(2011年)

https://doi.org/10.6342/NTU.2011.01201

摘要

在這篇論文裡，首先介紹了一種非常有用的區域特徵，稱為最穩定極值區域。這個特徵區域比同類的特徵子上絕大多數表現得很突出，並且有運算速度快的優點。爾後我們介紹了兩種不同的描述子去描述特徵區域，分別是利用共變矩陣正規化以及建立仿射不變特徵。仿射不變特徵是由特定的點所構成，比如說區域的幾何重心，對於凹面的切線，在凹面中對於切線最遠的點，在整個邊界中對於切點最遠的點。並且詳細了介紹凹面切點的演算法以及應用在最穩定極值區域的資料結構演算法。最後利用仿射不變特徵對於影像資料進行比對並評估效能。

關鍵字

最穩定極值區域 ； 仿射不變特徵 ； 共變異矩陣 ； 正規化 ； 公切線

並列摘要

At first, this paper introduced a powerful region feature – Maximally Stable Extremal Region (MSER), which has a better performance comparing to other region features. Then we use two different descriptors to describe the region, such as ellipse expression and local affine frame construction. Ellipses for each region are computed from covariant matrix and can be normalized to a circle. Local affine frame (LAF) is another description to feature region. In this paper we use several distinguished points, such as geometric center of the region, bi-tangent points of the region, the deepest points of the concavity and the farthest from the bi-tangent line. We also explain algorithms to MSER、LAF, including union-find and bin-sort for MSER detection and contour tracing, tangent from point to any polygon and steps to find bi-tangent points to construct LAF. Finally we compare the MSER to the other region detector and use MSER with LAF to match images from different deformation with image.

並列關鍵字

Maximally Stable Extremal Region ； Local Affine Frames (LAF) ； Covariance Matrix ； Normalization ； Bi-tangent line

參考文獻

[1] Ferrari, V., Tuytelaars, T., and Van Gool, L. 2001. Simultaneous object recognition and segmentation by image exploration. In Proceedings European Conference on Computer Vision, Prague, Czech Republic, pp. 40–54.

[4] Lowe, D. 2004. Distinctive image features from scale-invariant keypoints. International Journal on Computer Vision 60(2):91–110.

[5] Matas, J. Chum, O., Urban, M., and Pajdla, T. 2002. Robust wide-baseline stereo from maximally stable extremal regions. In Proceedings of the British Machine Vision Conference, Cardiff, UK, pp. 384–393.

[6] Matas, J., Chum, O., Urban, M., and Pajdla, T. 2004. Robust wide-baseline stereo from maximally stable extremal regions. Image and Vision Computing 22(10):761–767.

[7] Obdrˇz’alek, ˆ S. and Matas, J. 2002. Object recognition using local affine frames on distinguished regions. In Proceedings of the British Machine Vision Conference, Cardiff, UK, pp. 113–122.