A Closed-Form Solution to Natural Image Matting

By Anat Levin, Dani Lischinski, Yair Weiss. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2006

2,148 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/23a2d40bfea0995899b11f83c719313cb5ae1766

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

Semantic Object Selection (2014, methodology): “Matting methods [5, 14, 27] require a trimap to be specified.”
The Ignorant Led by the Blind: A Hybrid Human–Machine Vision System for Fine-Grained Categorization (2014, methodology): “…(Donahue and Grauman 2011), and runtime interactive computer vision systems for segmentation and tracking (Wu and Yang 2006; Rother et al. 2004; Levin et al. 2007; Vondrick et al. 2010) are all interesting related lines of research that apply to applications that are not considered in this…”
Automatic Red-Channel underwater image restoration (2015, methodology): “This method is faster and has no noticeable loss of detail with respect to the method based on the image matting technique (21), originally used in (17).”
Learning Guided Convolutional Network for Depth Completion (2019, methodology): “The kernel weights W ij are computed according to the guidance image I and a hand-crafted closed-form equation similar to the matting Laplacian from [49].”
Strong supervision from weak annotation: Interactive training of deformable part models (2011, background): “Active Labeling: Interactive labeling–also called active labeling–has been applied to interactive image segmentation or matting [17, 27, 13], semi-automated video annotation [31, 23, 1], and active classification for class-attribute models [4, 14].”
Single Image Dehazing by Multi-Scale Fusion (2013, methodology): “To refine the final depth-map, the transmission map values are extrapolated into the unknown regions, by a relatively computationally expensive matting strategy [24].”
Single Image Haze Removal Using Dark Channel Prior (2011, methodology): “Therefore, we can apply a closed-form framework of matting [16] to refine the transmission.”
A Closed-form Solution to Photorealistic Image Stylization (2018, methodology): “To avoid selecting one global scaling hyper-parameter, we resort to the matting affinity [13, 39] where the affinity between two pixels is based on means and variances of pixels in a local window.”
Real-Time High-Resolution Background Matting (2020, background): “Traditional (non-learning based) matting algorithms [6, 16, 17, 29, 9, 2, 7] require manual annotation (a trimap) and solve for the alpha matte in the ‘unknown’ region of the trimap.”
User-Friendly Interactive Image Segmentation Through Unified Combinatorial User Inputs (2010, methodology): “while soft segmentation algorithms such as [3], [4] extract a fractional (fuzzy) matte for an image.”
Initial results in underwater single image dehazing (2010, methodology): “He et al note that a closed form solution for natural image matting [12], can be used to refine a coarse depth prior.”
Single Image Motion Deblurring Using Transparency (2007, methodology): “We do not propose a new method to compute the transparency since there have been several successful natural image matting techniques proposed [3, 22, 14] which estimate foreground, background, and the alpha matte from a single input natural image.”
Single-image shadow detection and removal using paired regions (2011, methodology): “To evaluate our shadow detection and removal, we propose a new dataset with 108 natural scenes, in which ground truth is determined by taking two photographs of a scene after manipulating the shadows (either by blocking the direct light source or by casting a shadow into the image).”
Optimized Color Sampling for Robust Matting (2007, methodology): “We compare the performances of 7 algorithms: Bayesian matting [1], Belief Propagation matting [14], Global Poisson matting [12], Random Walk matting [5], KnockOut 2 [2], closed-form matting [7], and the robust algorithm proposed in this paper.”
Deep Image Matting (2017, methodology): “[8] take the matting results of [22] and [5] and normalized RGB colors as inputs and learn an end-to-end deep network to predict a new alpha matte.”
A Fast Single Image Haze Removal Algorithm Using Color Attenuation Prior (2015, methodology): “[46] replace the time-consuming soft matting [44] with standard median filtering, “median of median filter”, guided joint bilateral filtering [37]–[42] and guided image filtering, respectively.”
ScribbleSup: Scribble-Supervised Convolutional Networks for Semantic Segmentation (2016, methodology): “Formulation in this form is widely used for interactive image segmentation [4, 26, 17, 11, 16, 19].”
Video SnapCut: robust video object cutout using localized classifiers (2009, methodology): “…been shown to be more accurate than other matting affinities [Wang and Cohen 2007a], andLm is the matting Laplacian matrix defined in Equation 5 in [Levin et al. 2008]. λS is a weight to balance the data term and smoothness term, which is typically set to be 20 in our experiments, and can be…”