Comparison of Edge Detector Performance through Use in an Object Recognition Task

doi:10.1006/cviu.2001.0932

Computer Vision and Image Understanding

Volume 84, Issue 1, October 2001, Pages 160-178

https://doi.org/10.1006/cviu.2001.0932 Get rights and content

Abstract

This paper presents an empirical evaluation methodology for edge detectors. Edge detector performance is measured using a particular edge-based object recognition algorithm as a “higher-level” task. A detector's performance is ranked according to the object recognition performance that it generates. We have used a challenging train and test dataset containing 110 images of jeep-like images. Six edge detectors are compared and results suggest that (1) the SUSAN edge detector performs best and (2) the ranking of various edge detectors is different from that found in other evaluations.

References (16)

F. Bergholm
Edge focusing
IEEE Trans. Pattern Anal. Mach. Intell.
(1987)
M. Black et al.
Robust anisotropic diffusion
IEEE Trans. Image Process.
(1998)
J. Canny
A computational approach to edge detection
IEEE Trans. Pattern Anal. Mach. Intell.
(1986)
K. W. Bowyer, Christine, Kranenburg, and, Sean, Dougherty, Edge detector evaluation using empirical ROC curves, Comput....
K.I. Chang et al.
Evaluation of texture segmentation algorithms
Proceedings of IEEE Conference on Computer Vision and Pattern Recognition
(1999)
H. Christensen et al.
Guest Eds., Special Issue on Performance Evaluation
Machine Vision and Applications
(1997)
M. Heath et al.
A robust visual method for assessing the relative performance of edge detection algorithms
IEEE Trans. Pattern Anal. Mach. Intell.
(1997)

There are more references available in the full text version of this article.

Cited by (69)

Adaptive Cubic Spline Interpolation in CIELAB Color Space for Underwater Image Enhancement
2020, Procedia Computer Science
Underwater images suffer from haze, blue/green tint, and color distortion, caused by attenuation and scattering of light. Existing methods for restoration depend on transmission maps to inverse the effects based on the underwater image formation model, and enhancement methods often result in unnatural colors. In this paper, we propose a method to adaptively estimate a color correction curve in the CIE L*a*b* color space for single image enhancement. Haze and blue-green tint in underwater images are removed through this process. We demonstrate improved canny edge detection on underwater images through enhancement. Results are evaluated and compared with state-of-art methods using standard underwater image datasets of coral reefs with marine populations and man-made concrete structures. Enhancement of underwater images finds its use in areas such as marine habitat mapping, improvement of 3D reconstruction of underwater scenes, and visual studies of marine life.
Extraction and analysis of geological lineaments combining a DEM and remote sensing images from the northern Baoji loess area
2018, Advances in Space Research
Geological lineaments are important reactant of geological structure on the surface, and control the distribution of regional groundwater, geohazards, geothermal and earthquakes. The technology of geological lineaments extraction is of great significance for the analysis of regional plate movement and ore forming prognosis. However, the traditional methods are mainly based on the semi-automatic or manual visual interpretation, which consuming time and labor for the dependence on the rich professional experience and knowledge of the interpreter experts, and involving less in the special geomorphological regions. Taking Loess areas in northern Baoji as an example, this paper proposed a lineaments extraction algorithm based on tensor voting coupled Hough Transform, with the help of DEM and Landsat 8 OLI remote sensing images. Firstly, the best independent band combination of Landsat 8 OLI images for lineaments extraction were selected by using principal component analysis. Secondly, Gaussian high-pass filter was applied to sharpen the edge in the DEM and composite Landsat 8 OLI images. Linear boundary was extracted by tensor voting according to the conspicuousness of the vector sum superposition feature based on edge points. Finally, the Hough Transform was employed to search the edges and extracted the geological lineaments in this region. The experimental results showed that the orientation of lineaments was dominated by NW-SE and NE- SW, supplemented by NNW-SSE. Under the influence of the uplift of the eastern foot of Liupan Mountains and the southern margin of Ordos, the lineaments were mainly distributed over linear landforms, which were better consistent with previous studies about the direction of tectonics in this region. Compared with the segment tracing algorithm, this method has more applicability, efficiency, high practical value and scientific significance in the analysis of tectonic movement and evolution in special landform area.
Edge detection using G-lets based on matrix factorization by group representations
2017, Pattern Recognition
Citation Excerpt :
The irreducible representations have been used for image processing as convolution kernels by Lenz [22] who has also proposed an edge detection algorithm based on the kernel. More than improvements on the edge operators, the standard operators have been compared in different ways by many researchers [1,6,7,18–21] where the challenges of edge detection are discussed. Edge focusing is discussed in [2].
A new edge detection technique using transformation groups based G-lets filters is proposed in this paper. Discretizing gradients seem to produce discontinuity in classic edge detectors. No particular filter is capable of identifying meaningful edges at all scales and it increases computations with a multiscale approach. It is a challenge to get localized edges without spurious ones due to noise and integrate the obtained edges into meaningful object boundaries. Without breaking edge continuity and strictly localizing edges requires that filters do not blur the image during preprocessing. G-lets filters are found to be capable of performing well in most type of images including natural, noisy, low resolution and synthetic. In this paper, an edge detection algorithm using G-lets filters which are built by direct factorization of linear transformation matrices using irreducible representations is proposed. A multiresolution approach is shown to enhance the possibility of detecting faint edges. An edge tracing algorithm is presented to produce the edge image. The computational cost involved is comparatively lesser than existing filters. It is found that the geometries in the original image are preserved in the edge image. The edge tracing algorithm is capable of constructing object boundaries without the inner textures in a way that is not completely dependent on intensity thresholding. G-lets filters and the edge operator is found to be a promising algorithm for drastically bringing down the computations needed for realtime applications. The results are compared with BSDS500 boundary detection dataset using pb and global pb detectors.
Fast SUSAN edge detector by adapting step-size
2013, Optik
The SUSAN algorithm is a popular edge detector for its high localization precision, noise-robustness and good connectivity at junctions. Straightforward implementation of SUSAN edge detector is performed by sliding a circular mask pixel by pixel on an image. For each pixel in the raster scan order, neighbor pixels in the circular mask around it are taken and used to compute the USAN area, according to which whether the current pixel is an edge point or not is judged. However, redundant computation exists in the direct implementation of SUSAN edge detector because of the pixel-by-pixel sliding way of the circular mask, which makes it somewhat inefficient. In this paper, we propose a fast implementation of SUSAN edge detector to decrease the redundancy. Instead of sliding the mask by one pixel every time, we propose an adaptive type of raster scan referred to as adaptive sliding mask which adjusts the sliding step-size adaptively according to USAN area. Experiments show that the proposed method can speed up the SUSAN edge detector effectively without spoiling its detection performance.
Improved authentication and computation of medical data transmission in the secure IoT using hyperelliptic curve cryptography
2022, Journal of Supercomputing
An advanced approach to detect edges of digital images for image segmentation
2020, Applications of Advanced Machine Intelligence in Computer Vision and Object Recognition: Emerging Research and Opportunities