A fast and robust ellipse detector based on top-down least-square fitting

Yongtao Wang, Zheqi He, Xicheng Liu, Zhi Tang, Luyuan Li
2015 Procedings of the British Machine Vision Conference 2015  
Figure 1: Detecting ellipses from images Detecting ellipses efficiently and accurately in digital images is an fundamental problem to the field of pattern recognition and computer vision. Ellipse detection can be used in many applications. The existing algorithms such as [4] often use a bottom-up strategy to combine edge points or elliptical arcs into ellipses, which limits their robustness. In this paper, we propose a novel algorithm which can efficiently and accurately detect ellipses in
more » ... al images with a novel top-down scheme. The main idea of the proposed algorithm is to exploit a novel top-down fitting strategy to combine edge points into ellipses and use integral chain to speed up the fitting process. The proposed method is very efficient, which is faster than the most efficient method [1] which has been reported. Our experimental results also demonstrate that our algorithm is more robust than the state-of-the-art methods including [1] and ELSD [4].As shown in Fig. 2 , our method consists of four major steps. Figure 2: Workflow of the proposed ellipse detection method In the first step, we extract edge segments from the input image and then obtain line segments within each edge segment. We directly use the edge segment extraction method and the line segment detection method proposed in [2, 3] respectively, which are very computational efficient. Secondly, we detect candidate ellipses from each edge segment by the topdown analysis. In this step, we use the 1-D case(which is called "integral chain" hereafter) of integral image technique [5] to speed up the proposed ellipse detection algorithm.
doi:10.5244/c.29.156 dblp:conf/bmvc/WangHLTL15 fatcat:dqtoyg7gojc3ri4baaygiwubwu