Interactive Image Segmentation Using an Adaptive GMMRF Model

By Andrew Blake, Carsten Rother, Matthew A. Brown, Patrick Pérez, Philip H. S. Torr. 2004

646 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

Extreme Clicking for Efficient Object Annotation (2017, methodology): “Usually, this penalty depends on the RGB difference between pixels, being smaller in regions of high contrast [8, 6, 27, 42, 51, 64].”
Scene Text Recognition using Higher Order Language Priors (2009, methodology): “The unary potential yi(xi) gives the cost of the assignment: Xi = xi. Cost functions based on colour, location, and texton features have been commonly used for object segmentation [ 4 , 17, 25].”
How to Evaluate Foreground Maps (2014, background): “, salient object detection [10], object segmentation [11], and foreground-extraction [6].”
Decision tree fields (2011, methodology): “For pairwise interactions we generalize simple CRFs with contrast-sensitive pairwise potentials such as in [5, 4], the GrabCut system [24], and TextonBoost [29].”
A survey of graph theoretical approaches to image segmentation (2013, methodology): “[68] suggested using a Gaussian Mixture model (GMM) to approximate the distributions.”
Training Object Class Detectors from Eye Tracking Data (2014, neutral): “In traditional work using similar energy models [4,38,49], estimating the appearance models requires user interaction to indicate the image region containing the object (typically a manually drawn bounding-box or scribbles).”
User-Friendly Interactive Image Segmentation Through Unified Combinatorial User Inputs (2010, methodology): “where “misclassified pixels” excludes those from the unclassified region of the expert trimap [11], [26].”
Figure-ground segmentation by transferring window masks (2012, methodology): “For this, all masks in Nk are resized to the size of Wk in both their width and height dimensions.”
Learning Efficient Random Maximum A-Posteriori Predictors with Non-Decomposable Loss Functions (2013, methodology): “A popular approach for segmentation is the GrabCut approach [2, 1].”
Learning to Detect A Salient Object (2007, neutral): “β is a robust parameter that weights the color contrast, and can be set as β = (2〈||Ix−Ix′ ||(2)〉) [1], where 〈·〉 is the expectation operator.”
Graph Cuts and Efficient N-D Image Segmentation (2006, methodology): “…cut framework for object segmentation was first proposed in Boykov and Jolly (2001) and recently further developed in (Boykov and Kolmogorov, 2003; Blake et al., 2004; Rother et al., 2004; Li et al., 2004; Kirsanov and Gortler, 2004; Li et al., 2005; Wang et al., 2005; Kolmogorov et al., 2005;…”
TextonBoost for Image Understanding: Multi-Class Object Recognition and Segmentation by Jointly Modeling Texture, Layout, and Context (2007, background): “Individual objects tend to have a relatively compact color distribution, and exploiting an accurate instance color model can therefore dramatically improve segmentation results (Blake et al. 2004).”
Observing Human-Object Interactions: Using Spatial and Functional Compatibility for Recognition (2009, methodology): “We sample pixels from training images of the manipulable objects and build a 3D model in the RGB space. pModelðr; g; bÞ ¼ 1 N XN i¼1 K rðr riÞK gðg giÞK bðb biÞ: ð7Þ Given a test image, we first use the shape-based classifier to detect potential object candidates.”
Robust Higher Order Potentials for Enforcing Label Consistency (2008, background): “This problem can be overcome by using sophisticated potential functions based on colour, texture, location, and shape priors as shown by [1, 5, 16, 25, 29].”
Multiview Stereo via Volumetric Graph-Cuts and Occlusion Robust Photo-Consistency (2007, methodology): “A probabilistic formu lation of interactive segmentation with a more elaborate foreground/background model was give n in Blakeet al [1].”
OBJ CUT (2005, methodology): “We cast the problem of object category specific segmentation as that of estimating an MRF (representing bottom up information) which is influenced by a set of latent variables, the PS (representing top down information), encouraging the MRF to resemble the object.”
GeoS: Geodesic Image Segmentation (2008, background): “More iterations as in [26] help reduce the amount of manual interaction.”