Learning affinity functions for image segmentation: combining patch-based and gradient-based approaches

By Charless C. Fowlkes, David R. Martin, Jitendra Malik. 2003

165 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition

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

Citations in the corpus

Contour Detection and Hierarchical Image Segmentation (2011, methodology): “As input to the spectral clustering stage, we construct a sparse symmetric affinity matrix W using the intervening contour cue [49], [64], [65], the maximal value of mPb along a line connecting two pixels.”
Learning a classification model for segmentation (2003, methodology): “This framework connects the segmentation as a computational problem with the ecological statistics of na tural images [9, 2] and the rich theory of learning.”
Maximin affinity learning of image segmentation (2009, background): “where the sum is over all nearest neighbor pixel pairs 〈i, j〉 and c is the number of nearest neighbors [9].”
Affinity CNN: Learning Pixel-Centric Pairwise Relations for Figure/Ground Embedding (2015, background): “For example, image segmentation follows from edge detection [12, 1, 2, 7, 4]; figure/ground, occlusion, or depth layering follows from reasoning over discrete contours or regions [27, 16, 21, 36, 18] with some systems also reliant on motion cues [30, 15, 32, 31].”
A Statistical Approach to Material Classification Using Image Patch Exemplars (2009, background): “In particular, [16], [32], [40] have all noted that, in their segmentation and classification tasks, filters with small support have outperformed the same filters at larger scales.”
Constrained parametric min-cuts for automatic object segmentation (2010, methodology): “Learning methods have been applied for segmentation, to distinguish real from apparent contours [13, 10, 16], similar from dissimilar superpixels [15], and good from bad regions [25].”
Measuring and Predicting Object Importance (2011, background): “However, our scene photographs are large and complex, and, in our hands, segmentations produced by state of the art algorithms (Fowlkes et al. 2003; Rabinovich et al. 2006) are not as detailed as the verbal responses.”
Learning full pairwise affinities for spectral segmentation (2010, background): “Therefore, to produce high-quality segmentation results with object-level details, it is important to define the graph affinities gained by integrating local grouping cues.”
Toward Objective Evaluation of Image Segmentation Algorithms (2007, background): “Information-based: Work in [6], [9] proposes to formulate the problem as that of evaluating an affinity function that gives the probability of two pixels belonging to the same segment.”
From contours to regions: An empirical evaluation (2009, methodology): “From mPb, we define an affinity matrix W between pixels using the intervening contour cue [18].”
Comparison of segmentation algorithms for fluorescence microscopy images of cells (2011, methodology): “TET and TEE are very similar to precision and recall from precision-recall framework used to evaluate boundary-based segmentations of natural scenes (40–42).”
Convolutional Networks Can Learn to Generate Affinity Graphs for Image Segmentation (2010, methodology): “Much prior work has been conﬁned to estimating afﬁnities by classifying carefully chosen hand-designed image features, which are often image domain speciﬁc (Fowlkes et al., 2003).”
Efficient Multilevel Brain Tumor Segmentation With Integrated Bayesian Model Classification (2008, unknown): “The parameters are fixed and predetermined through some heuristic techniques or learned from training data [37].”
Deep clustering: Discriminative embeddings for segmentation and separation (2015, methodology): “Although affinity-based methods were originally unsupervised inference methods, multiple-kernel learning methods such as [17, 18] were later introduced to train weights used to combine separate affinity measures.”
Machines that learn to segment images: a crucial technology for connectomics. (2010, methodology): “, 2010) as well as natural images (Fowlkes et al., 2003).”
Boundary Learning by Optimization with Topological Constraints (2010, background): “Finally, it should also be possible to extend BLOTC to boundary detectors that classify the edges of an affinity graph rather than pixels [28, 29, 9].”
CPMC: Automatic Object Segmentation Using Constrained Parametric Min-Cuts (2012, background): “Methods that learn low-level statistics have been applied to distinguish real from apparent contours [41], [42], [43] and similar to dissimilar superpixels [25].”
Object detection and segmentation from joint embedding of parts and pixels (2011, background): “Objects can also occlude one another, suggesting a continuous measure of figure/ground ordering is most appropriate in capturing this aspect of scene layout.”
Shared Segmentation of Natural Scenes Using Dependent Pitman-Yor Processes (2008, methodology): “We compare our results to the normalized cuts spectral clustering method with varying numbers of segments (NCut(K)), and a high-quality affinity function based on color, texture, and intervening contour cues [13].”