Geodesic graph cut for interactive image segmentation

By Brian L. Price, Bryan S. Morse, Scott D. Cohen. 2010

211 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): “To reduce annotation time many interactive segmentation techniques have been proposed, which require the user to input either a bounding box around the object [42, 51, 71], or scribbles [3, 16, 24, 26, 27, 42, 50, 64, 65, 72], or clicks [31, 69].”
Deep Interactive Object Selection (2016, methodology): “Strokebased methods are popular, and use a number of underlying algorithms, including normalized cuts [22], graph cut [2, 11, 26], geodesics [1, 4], the combination of graph cut and geodesics [20, 9] and random walks [8].”
Guide Me: Interacting with Deep Networks (2018, background): “Prior to deep learning, several systems have been proposed for semi-automatic segmentation, that allowed the user to interfere with the result or to provide hints to the system via seed points [8, 52], bounding boxes [24, 37, 54], contours or scribbles [7, 23, 58], eye gaze [26] or in the form of bi”
Content-Aware Multi-Level Guidance for Interactive Instance Segmentation (2019, methodology): “More recent methods are based on graph cuts [7, 45, 49, 30], geodesics [5, 13], and or a combination of the two [19, 44].”
Y Qualification Main task Y Return N N Y User qualified ? N Feedback Pass ? Pass ? Annotator training Annotating images (2017, methodology): “To reduce annotation time many interactive segmentation techniques have been proposed, which require the user to input either a bounding box around the object [42, 51, 71], or scribbles [3, 16, 24, 26, 27, 42, 50, 64, 65, 72], or clicks [31, 69].”
PhraseClick: Toward Achieving Flexible Interactive Segmentation by Phrase and Click (2020, methodology): “For example, [5,31,49] adopt graphical models, [18] employs random walker and [3, 11, 47] are based on geodesic approaches.”
Saliency-aware geodesic video object segmentation (2015, methodology): “In this paper, we try to introduce geodesic distance into our totally automatic segmentation framework, which is different with previous approaches [3, 25, 2, 10] that require careful user assistance.”
MultiSeg: Semantically Meaningful, Scale-Diverse Segmentations From Minimal User Input (2019, methodology): “Interactive Image Segmentation: Early interactive image segmentation methods mainly exploit boundary properties for segmentation [17, 29] while more recent approaches are based on graphical models, such as graph cut [4, 21, 33], random walker [11] and geodesic approaches [2, 7, 31].”
DeepCut: Object Segmentation from Bounding Box Annotations using Convolutional Neural Networks (2016, background): “active [6], [8], [13], [14] and fully automated segmentation”
Ilastik: Interactive learning and segmentation toolkit (2011, methodology): “Methods based on graph cut [5, 8, 4] extend this idea by explicitly modeling a boundary component.”
MILCut: A Sweeping Line Multiple Instance Learning Paradigm for Interactive Image Segmentation (2014, methodology): “The GrabCut dataset has been used as a benchmark in multiple previous works [9, 20, 28, 39].”
DenseCut: Densely Connected CRFs for Realtime GrabCut (2015, methodology): “Following [LSTS04, PMC10], we use the foreground/background term of the form P ( x i ) = P ( Θ xi , I i ) P ( Θ 0 , I i )+ P ( Θ 1 , I i ) , where P ( Θ 0 , I i ) , P ( Θ 1 , I i ) ∈ ( 0 , ∞ ) represent the probability density value of a pixel color I i belonging to the background color model Θ 0…”
Saliency-Aware Video Object Segmentation (2018, background): “The geodesic distance, which has been shown to be effective in many interactive computer vision applications [16], [17], [18], [19], has the power of abstracting object structure to efﬁciently determine its central regions by assigning higher saliency values to more representative regions.”
Geodesic Object Proposals (2014, background): “However, the zero level set is sensitive to the detailed form of the masks and may not adhere to object boundaries [18].”
A Fully Convolutional Two-Stream Fusion Network for Interactive Image Segmentation (2018, background): “To take the advantage of both graph cut and geodesic distance, Price et al. [4] propose geodesic graph cut which incorporates geodesic distance into a graph cut-based framework.”
Regional Interactive Image Segmentation Networks (2017, unknown): “To achieve this, many algorithms have been proposed in the literature, such as graph cut [4, 22, 16], random walker [11], geodesic segmentation [1, 6], the combination of graph cut and geodesics [21], growcut [26], etc.”
Interactive Image Segmentation with Latent Diversity (2018, methodology): “Until recently, prior knowledge about object appearance was encoded primarily in hand-crafted features and distance metrics [4, 37, 12, 2, 9, 13, 35, 5, 24, 8].”
Beyond pixels: A comprehensive survey from bottom-up to semantic image segmentation and cosegmentation (2015, background): “Then the geodesic distance can be incorporated as a kind of data term in the energy function as in [80].”
Continuous Adaptation for Interactive Object Segmentation by Learning from Corrections (2019, methodology): “[46], DAVIS [40] and Berkeley [37]. 2. Related Work Interactive Object Segmentation. Traditional methods model interactive segmentation as an energy minimization problem on a graph deﬁned over pixels [12,46,7,24,41]. User inputs are used to create an image-speciﬁc appearance model based on low-level”
Normalized Cut Loss for Weakly-Supervised CNN Segmentation (2018, methodology): “It is true that many off-line interactive segmentation algorithms exist [42, 31, 40].”