Evaluating Edge Detection through Boundary Detection
- PDF / 2,561,118 Bytes
- 15 Pages / 600.03 x 792 pts Page_size
- 66 Downloads / 207 Views
Evaluating Edge Detection through Boundary Detection Song Wang, Feng Ge, and Tiecheng Liu Department of Computer Science and Engineering, University of South Carolina, Columbia, SC 29208, USA Received 27 February 2005; Revised 6 June 2005; Accepted 30 June 2005 Edge detection has been widely used in computer vision and image processing. However, the performance evaluation of the edgedetection results is still a challenging problem. A major dilemma in edge-detection evaluation is the difficulty to balance the objectivity and generality: a general-purpose edge-detection evaluation independent of specific applications is usually not well defined, while an evaluation on a specific application has weak generality. Aiming at addressing this dilemma, this paper presents new evaluation methodology and a framework in which edge detection is evaluated through boundary detection, that is, the likelihood of retrieving the full object boundaries from this edge-detection output. Such a likelihood, we believe, reflects the performance of edge detection in many applications since boundary detection is the direct and natural goal of edge detection. In this framework, we use the newly developed ratio-contour algorithm to group the detected edges into closed boundaries. We also collect a large data set (1030) of real images with unambiguous ground-truth boundaries for evaluation. Five edge detectors (Sobel, LoG, Canny, Rothwell, and Edison) are evaluated in this paper and we find that the current edge-detection performance still has scope for improvement by choosing appropriate detectors and detector parameters. Copyright © 2006 Hindawi Publishing Corporation. All rights reserved.
1.
INTRODUCTION
Edge detection is a very important feature-extraction method that has been widely used in many computer vision and image processing applications. The basic idea of most available edge detectors is to locate some local object-boundary information in an image by thresholding and skeletonizing the pixel-intensity variation map. Since the earliest work by Julez [1] in 1959, a huge number of edge detectors has been developed from different perspectives (e.g., [2–9]). A very natural and important question is then: which edge detector and detector-parameter settings can produce better edgedetection results? This strongly motivates the development of a general and systematic way of evaluating the edge-detection results. Prior edge-detection evaluation methods can be categorized in several ways. First, they can be classified as subjective and objective methods. The former uses the humanvisual observation and decision to evaluate the performance of edge detection. Given the inherent inconsistency in human perception, subjective evaluation results may exhibit a large variance for different observers. In objective methods, quantitative measures are defined based solely on images and the edge-detection results. Second, edge-detection evaluation methods can be categorized according to their requirement of the ground truth. With the ground truth, edge detecti
Data Loading...
