## Mean Shift Analysis and Applications (1999)

### Cached

### Download Links

- [www.scr.siemens.com]
- [www.caip.rutgers.edu]
- [pages.cs.wisc.edu]
- DBLP

### Other Repositories/Bibliography

Citations: | 157 - 8 self |

### BibTeX

@MISC{Comaniciu99meanshift,

author = {Dorin Comaniciu and Peter Meer},

title = {Mean Shift Analysis and Applications},

year = {1999}

}

### Years of Citing Articles

### OpenURL

### Abstract

A nonparametric estimator of density gradient, the mean shift, is employed in the joint, spatial-range (value) domain of gray level and color images for discontinuity preserving filtering and image segmentation. Properties of the mean shift are reviewed and its convergence on lattices is proven. The proposed filtering method associates with each pixel in the image the closest local mode in the density distribution of the joint domain. Segmentation into a piecewise constant structure requires only one more step, fusion of the regions associated with nearby modes. The proposed technique has two parameters controlling the resolution in the spatial and range domains. Since convergence is guaranteed, the technique does not require the intervention of the user to stop the filtering at the desired image quality. Several examples, for gray and color images, show the versatilityofthe method and compare favorably with results described in the literature for the same images.

### Citations

2185 | Density Estimation for Statistics and Data Analysis - Silverman - 1986 |

1287 | Scale-space and edge detection using anisotropic diffusion
- Perona, Malik
- 1990
(Show Context)
Citation Context ...ed in [15] do not have a stopping criterion. After a su cientnumber of iterations, the processed image collapses to a at surface. The same observation is valid for other adaptive smoothing techniques =-=[9, 10]-=-. The process de ned by mean shift is run till convergence and maintains the structure of the data. 6 Segmentation The mean shift segmentation in the spatial-range domain has the same simple design as... |

629 | A.: Region Competition: Unifying Snakes, Region Growing, and Bayes/MDL for Multiband Image Segmentation - Zhu, Yuille - 1996 |

392 |
Mean shift, mode seeking, and clustering
- Cheng
- 1995
(Show Context)
Citation Context ...-4 -2 Figure 1: Successive computations of the mean shift define a path leading to a local density maximum. The convergence of the mean shift has been justified as a consequence of relation (7), (see =-=[2]-=-). However, while it is true that the mean shift vector M h (x) has the direction of the gradient of the density estimate at x, it is not apparent that the density estimate at locationssfy k g k=1;2::... |

370 | Multivariate density estimation - SCOTT - 1992 |

301 |
The estimation of the gradient of a density function, with applications in pattern recognition
- Fukunaga, Hostetler
- 1975
(Show Context)
Citation Context ...ge smoothing and for segmentation. The mean shift estimate of the gradient of a density function and the associated iterative procedure of mode seeking have been developed byFukunaga and Hostetler in =-=[6]-=-. Only recently, however, the nice properties of data compaction and dimensionality reduction of the mean shift have been exploited in low level computer vision tasks (color space analysis [3], face t... |

187 | Robust analysis of feature spaces: color image segmentation
- Comaniciu, Meer
- 1997
(Show Context)
Citation Context ...n [7]. Only recently, however, the nice properties of data compaction and dimensionality reduction of the mean shift have been exploited in low level computer vision tasks (e.g., color space analysis =-=[4]-=-, face tracking [1]). In this paper we describe a new application based on the theoretical results obtained in [5]. We show that high quality edge preserving filtering and image segmentation can be ob... |

145 | An adaptive clustering algorithm for image segmentation - Pappas - 1992 |

100 | Image segmentation using local variation - Felzenszwalb, Huttenlocher - 1998 |

98 |
G.: Adaptive smoothing: A general tool for early vision
- Saint-Marc, Chen, et al.
- 1991
(Show Context)
Citation Context ...ed in [15] do not have a stopping criterion. After a su cientnumber of iterations, the processed image collapses to a at surface. The same observation is valid for other adaptive smoothing techniques =-=[9, 10]-=-. The process de ned by mean shift is run till convergence and maintains the structure of the data. 6 Segmentation The mean shift segmentation in the spatial-range domain has the same simple design as... |

65 | Distribution free decomposition of multivariate data
- Comaniciu, Meer
- 1999
(Show Context)
Citation Context ...ft have been exploited in low level computer vision tasks (e.g., color space analysis [4], face tracking [1]). In this paper we describe a new application based on the theoretical results obtained in =-=[5]-=-. We show that high quality edge preserving filtering and image segmentation can be obtained by applying the mean shift in the combined spatial-range domain. The methods we developed are conceptually ... |

33 |
Real time face and object tracking as a component of a perceptual user interface
- Bradski
- 1998
(Show Context)
Citation Context ...y, however, the nice properties of data compaction and dimensionality reduction of the mean shift have been exploited in low level computer vision tasks (e.g., color space analysis [4], face tracking =-=[1]-=-). In this paper we describe a new application based on the theoretical results obtained in [5]. We show that high quality edge preserving filtering and image segmentation can be obtained by applying ... |

29 | Image Segmentation from Consensus Information - Cho, Meer - 1997 |

10 |
Linear Density Estimates
- Terrell
- 1990
(Show Context)
Citation Context ... paper is the adaptive de nition of the normalization constants. To takeinto account the nonstationarity of the input adaptive kernel estimation techniques were proposed in the statistical literature =-=[14]-=-, however for less complex data. Beside exploiting a priori information (often available for low level vision) robust image understanding methods can also be helpful. 5.3 Experiments Mean shift lterin... |

2 |
Time Face and Object Tracking as a Component ofaPerceptual User Interface
- Bradski, `'Real
- 1998
(Show Context)
Citation Context ...ecently, however, the nice properties of data compaction and dimensionality reduction of the mean shift have been exploited in low level computer vision tasks (color space analysis [3], face tracking =-=[1]-=-). In this paper we describe a new application based on the theoretical results obtained in [4]. We show that high quality edge preserving ltering and image segmentation can be obtained by applying th... |

2 |
Multiscale Image Segmentation byIntegrated Edge and Region Detection
- Tabb, Ahuja
- 1997
(Show Context)
Citation Context ...9 with those from [16, Figure 15] obtained through a complex global optimization. 6.2 Discussion It is interesting to contrast the mean shift segmentation with those based on the attraction force eld =-=[13]-=- and edge ow propagation [7]. While all the three methods employ avector eld to detect regions in the spatial domain, only the mean shift based segmentation has strong statistical foundations. Our met... |

1 |
Edge Flow: AFramework of Boundary Detection and Image Segmentation
- Ma, Manjunath
- 1997
(Show Context)
Citation Context ... 15] obtained through a complex global optimization. 6.2 Discussion It is interesting to contrast the mean shift segmentation with those based on the attraction force eld [13] and edge ow propagation =-=[7]-=-. While all the three methods employ avector eld to detect regions in the spatial domain, only the mean shift based segmentation has strong statistical foundations. Our method associates the current p... |

1 |
Bilateral Filtering for Gray and Color Images", Int'l Conf
- Tomasi, Manduchi
- 1998
(Show Context)
Citation Context ...s of the eyes and the whiskers remained crisp. 5.4 Comparison to Bilateral Filtering We note here two important di erences between the mean shift and bilateral ltering proposed by Tomasi and Manduchi =-=[15]-=-. Both methods are based on the same principle, the simultaneous processing of both the spatial and range domains. However, while the bilateral ltering uses a static window in the two domains, the mea... |