Statistical Edge Detection: Learning and Evaluating Edge Cues

By Scott Konishi, Alan L. Yuille, James M. Coughlan, Song-Chun Zhu. 2003

365 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Industrial and Manufacturing Engineering, Media Technology

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

In Industrial and Manufacturing Engineering

In Media Technology

Citations in the corpus

Semantic Image Segmentation with Task-Specific Edge Detection Using CNNs and a Discriminatively Trained Domain Transform (2015, background): “Edge detection The edge/contour detection task has a long history [25, 1, 11], which we will only briefly review.”
Pushing the Boundaries of Boundary Detection using Deep Learning (2015, background): “As in all works following the introduction of human-annotated datasets (Konishi et al., 2003; Martin et al., 2004), e.g. (Dollar et al., 2006; Arbelaez et al., 2011; Ren, 2008; Kokkinos, 2010a; Ren & Bo, 2012; Doll ´ ar & Zitnick, 2015), we use machine learning to optimize the performance of our…”
Computational analysis and learning for a biologically motivated model of boundary detection (2008, methodology): “In [47,54] the features and classifiers that practically proved to offer the best performance are used, using e.”
End-to-End Instance Edge Detection (2022, background): “The early 2000s saw approaches driven by information theory such as statistical edges [34], Pb [47] and gPB [2].”
Object perception as Bayesian inference. (2004, background): “Other workers have learned posterior probabilities p(edge|φ(I)) and p(not − edge|φ(I)) where φ(I) are local image features (Konishi et al 2003).”
Image Parsing: Unifying Segmentation, Detection, and Recognition (2003, neutral): “In principle, statistical edge detectors (Konishi et al., 2003) would be preferable to Canny because they give discriminative probabilities q(w j | Tst j (I)) learnt from training data.”
PASCAL Boundaries: A Class-Agnostic Semantic Boundary Dataset (2015, methodology): “abled the identiﬁcation of the fact that low-level ﬁltering techniques such as the Canny edge detector [5] were limited in many ways. Moreover, the ﬁrst real statistical techniques for edge detection [14,15] were developed and tested on these datasets. The Sowerby dataset being too small motivated M”
Ultrasound image segmentation: a survey (2006, methodology): “5The idea of statistical edge detection was originally proposed by Yuille; See [142] and the references therein, for a recent paper on this.”
Supervised Learning of Edges and Object Boundaries (2006, methodology): “Two other relevant algorithms are Pb , proposed by [11], and the method presented in [13].”
Semantic Edge Detection with Diverse Deep Supervision (2018, methodology): “rious ﬁlters (e.g. Sobel [24] and Canny [25]) or complex models [26], [27] to detect pixels with highest gradients in their local neighborhoods [28]–[30]. To the best of our knowledge, Konishi et al. [31] proposed the ﬁrst data-driven edge detector in which, unlike previous model based approaches, e”
Learning a Dictionary of Shape Epitomes with Applications to Image Labeling (2013, neutral): “Most existing edge detection systems such as [3, 4, 1] employ a pixel-based representation of image edges.”
Bayesian models of object perception. (2003, background): “We can then infer the positions of edges in an image by calculating edge-filter responses and evaluating the ratio Pon/Poff — the higher the ratio the more likely there is an edge [21 ].”
Detecting Object Boundaries Using Low-, Mid-, and High-level Information (2007, result): “1, suggests that these neighborhoods are too small, though better results might be obtained by using slightly bigger neighborhoods and explicit learning [9].”
DeepContour: A deep convolutional feature learned by positive-sharing loss for contour detection (2015, methodology): “[24], contour detection is formulated as a discrimination task specified by a likelihood ratio tested on the filter responses.”
Boundary Detection Using F-Measure-, Filter- and Feature- (F3) Boost (2010, background): “A revival of research on boundary detection has been observed during the last years, largely due to the introduction of ground-truth labeled datasets [3, 4] which facilitated the treatment of the problem in a machine learning framework, while weeding out many of the heuristics previously used in edg”
Edge and line oriented contour detection: State of the art (2011, methodology): “Examples are the Berkeley (300 images) [2], South Florida (50 images), Sowerby (100 images) [124], Corel (5000 images) [282] and RuG (40 images) [22] datasets.”