## Efficient Energies and Algorithms for Parametric Snakes (2004)

Venue: | IEEE Transactions on Image Processing |

Citations: | 22 - 7 self |

### BibTeX

@ARTICLE{Jacob04efficientenergies,

author = {Mathews Jacob and Thierry Blu and Michael Unser},

title = {Efficient Energies and Algorithms for Parametric Snakes},

journal = {IEEE Transactions on Image Processing},

year = {2004},

pages = {1231--1244}

}

### OpenURL

### Abstract

Abstract—Parametric active contour models are one of the preferred approaches for image segmentation because of their computational efficiency and simplicity. However, they have a few drawbacks which limit their performance. In this paper, we identify some of these problems and propose efficient solutions to get around them. The widely-used gradient magnitude-based energy is parameter dependent; its use will negatively affect the parametrization of the curve and, consequently, its stiffness. Hence, we introduce a new edge-based energy that is independent of the parameterization. It is also more robust since it takes into account the gradient direction as well. We express this energy term as a surface integral, thus unifying it naturally with the region-based schemes. The unified framework enables the user to tune the image energy to the application at hand. We show that parametric snakes can guarantee low curvature curves, but only if they are described in the curvilinear abscissa. Since normal curve evolution do not ensure constant arc-length, we propose a new internal energy term that will force this configuration. The curve evolution can sometimes give rise to closed loops in the contour, which will adversely interfere with the optimization algorithm. We propose a curve evolution scheme that prevents this condition. Index Terms—Active contour, curve, segmentation, snake, spline.

### Citations

3080 | Snakes: Active contour models
- Kass, Witkin, et al.
- 1988
(Show Context)
Citation Context ...ergy term. The popular curve representation schemes in the snake literature are as follows: 1) point-based snakes, where the curve is an ordered collection of discrete points (also termed as snaxels) =-=[3]-=-–[5]; 2) parametric snakes, where the curve is described continuously in a parametric form, using basis functions such as B-splines [6]–[9], Fourier exponentials [10], [11], etc.; Manuscript received ... |

1078 | M.: Fast geodesic active contours - Goldenberg, Kimmel, et al. - 2001 |

630 | Shape modeling with front propagation: A level set approach",Pattern Analysis and Machine Intelligence - Malladi, Sethian, et al. - 1995 |

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

491 | Snakes, shapes, and gradient vector flow - Xu, Prince - 1998 |

453 | Deformable models in medical image analysis: a survey
- McInerney, Terzopoulos
- 1996
(Show Context)
Citation Context ...omatic approach may be attributed to its ability to aid the segmentation process with a priori knowledge and user interaction. Extensive research in this area has resulted in many snake variants [1], =-=[2]-=-; these are distinguished mainly by the type of curve representation used and the choice of the image energy term. The popular curve representation schemes in the snake literature are as follows: 1) p... |

312 | O.: Statistical shape influence in geodesic active contours
- Leventon, Grimson, et al.
(Show Context)
Citation Context ...mputationally more complex since they evolve a surface rather than a curve. Also, since the curve representation is implicit, it is much more challenging to introduce shape priors into this framework =-=[21]-=-. In this paper, we focus on general parametric snakes due to its computational advantages and simplicity. We will start by taking a critical look at them, identifying some of their limitations and pr... |

292 | Use of active shape models for locating structures in medical images - Cootes, Hill, et al. - 1994 |

279 |
Finite-Element Methods for Active Contour Models and Balloons for 2-D and 3-D Images
- Cohen
- 1993
(Show Context)
Citation Context ...ages. The edge-based energy can give a good localization of the contour near the boundaries. Unfortunately, it has a small basin of attraction, thus requiring a good initialization or a balloon force =-=[32]-=-. On the other hand, the region-based energy have a large basin of attraction and can converge even if explicit edges are not present [25]. However, it does not give as good a localization as the edge... |

240 |
Geometric Partial Differential Equations and Image Analysis
- Sapiro
- 2001
(Show Context)
Citation Context ...extra bookkeeping to accommodate changes in topology. Geometric approaches offer great flexibility as far as the curve topology is considered; they presently constitute a very promising research area =-=[14]-=-–[16]. However, they tend to be computationally more complex since they evolve a surface rather than a curve. Also, since the curve representation is implicit, it is much more challenging to introduce... |

240 | Splines: A perfect fit for signal and image processing
- Unser
- 1999
(Show Context)
Citation Context ... parametrization is the curvilinear abscissa (i.e., the knots are uniformly spaced on the curve). The use of spline curves also brings in additional gains due to the existence of efficient algorithms =-=[35]-=-, the local control of the contour due to the finite support of the B-spline basis function and their good approximation properties [36]. Due to these nice properties, we choose cubic spline curves in... |

156 |
VESE L.: Active contours without edges
- CHAN
(Show Context)
Citation Context ...mage in the region bounded by the curve. B. Region-Based Image Energy Recent research in active contours is increasingly focusing on the use of statistical region-based image energy [11], [19], [24], =-=[25]-=-. This type of energy can provide the snake with vital boundary information, especially while it is far away from the real contour, thus resulting in a larger basin of attraction. The use of this ener... |

100 |
Deformable boundary finding in medical images by integrating gradient and region information
- Chakraborty, Staib, et al.
- 1996
(Show Context)
Citation Context ...nts (also termed as snaxels) [3]–[5]; 2) parametric snakes, where the curve is described continuously in a parametric form, using basis functions such as B-splines [6]–[9], Fourier exponentials [10], =-=[11]-=-, etc.; Manuscript received June 20, 2003; revised January 7, 2004. This work was supported by the Swiss National Science Foundation under Grant 2100-053 540. The associate editor coordinating the rev... |

69 | Image segmentation using active contours: Calculus of variation or shape optimization
- Aubert, Barlaud, et al.
(Show Context)
Citation Context ...therefore proportional to . 2) Partial derivatives of the unified image energy. For closed curves, we preferentially use the unified energy to optimize the curve. In line with the work of [10], [11], =-=[37]-=-, we now use Green’s Theorem (12) to convert region integrals (over the region bounded by a closed curve) to integrals over the curve; our main motivation is computational efficiency. (20) can be effi... |

68 |
Statistical region snake-based segmentation adapted to different physical noise models
- Chesnaud, Refregier, et al.
- 1999
(Show Context)
Citation Context ... the image in the region bounded by the curve. B. Region-Based Image Energy Recent research in active contours is increasingly focusing on the use of statistical region-based image energy [11], [19], =-=[24]-=-, [25]. This type of energy can provide the snake with vital boundary information, especially while it is far away from the real contour, thus resulting in a larger basin of attraction. The use of thi... |

64 | Geodesic active regions for supervised texture segmentation
- Paragios, Deriche
- 1999
(Show Context)
Citation Context ...For a -D vector image ,wehave (19) Note that the region information from the vector data is efficiently concatenated into the scalar image . This framework is used for the segmentation of textures in =-=[31]-=-. They obtain an appropriate vector image from the gray level image using a Gabor filterbank. C. Unified Image Energy Both of the above-mentioned energies (edge-based and region-based) have their own ... |

64 | Quantitative fourier analysis of approximation techniques: part i–interpolations and projections
- Blu, Unser
- 1999
(Show Context)
Citation Context ...n additional gains due to the existence of efficient algorithms [35], the local control of the contour due to the finite support of the B-spline basis function and their good approximation properties =-=[36]-=-. Due to these nice properties, we choose cubic spline curves in our implementation. However, the theory we present in this paper is general enough to accommodate for any other representation in terms... |

61 | B-Spline Snakes: A Flexible Tool for Parametric Contour Detection
- Brigger, Hoeg, et al.
- 2000
(Show Context)
Citation Context ...dered collection of discrete points (also termed as snaxels) [3]–[5]; 2) parametric snakes, where the curve is described continuously in a parametric form, using basis functions such as B-splines [6]–=-=[9]-=-, Fourier exponentials [10], [11], etc.; Manuscript received June 20, 2003; revised January 7, 2004. This work was supported by the Swiss National Science Foundation under Grant 2100-053 540. The asso... |

58 |
Spline functions and the problem of graduation
- Schoenberg
- 1964
(Show Context)
Citation Context ...uch more expensive. B. Choice of the Scaling Basis Function As mentioned before, the parametric representations can guarantee a small value of . Using the well-known variational properties of splines =-=[34]-=-, we can show that the minimization of subject to interpolation constraints yields a cubic spline curve with knots at the integers. Thus, the cubic B-spline model appears to be the most natural choice... |

57 | Hierarchical Active Shape Models, Using the Wavelet Transform
- Davatzikos, Tao, et al.
(Show Context)
Citation Context ...nstraints to ensure smoothness [9], [17]. Since both the above mentioned schemes represent the curve explicitly, it is easy to introduce a priori shape constraints into the snake framework [10], [18]–=-=[20]-=-. It is also straightforward to accommodate user interaction; this is often done by allowing the user to specify points through which the curve should go through [3]. However, these models offer less ... |

56 |
Bsnakes: implementation and applications to stereo
- Menet, Saint-Marc, et al.
- 1990
(Show Context)
Citation Context ...n ordered collection of discrete points (also termed as snaxels) [3]–[5]; 2) parametric snakes, where the curve is described continuously in a parametric form, using basis functions such as B-splines =-=[6]-=-–[9], Fourier exponentials [10], [11], etc.; Manuscript received June 20, 2003; revised January 7, 2004. This work was supported by the Swiss National Science Foundation under Grant 2100-053 540. The ... |

47 |
Deformable template models: A review
- Jain, Zhong, et al.
- 1998
(Show Context)
Citation Context ...miautomatic approach may be attributed to its ability to aid the segmentation process with a priori knowledge and user interaction. Extensive research in this area has resulted in many snake variants =-=[1]-=-, [2]; these are distinguished mainly by the type of curve representation used and the choice of the image energy term. The popular curve representation schemes in the snake literature are as follows:... |

47 | Design of steerable filters for feature detection using canny-like criteria
- Jacob, Unser
- 2004
(Show Context)
Citation Context ...a generalized form of (7) by substituting with other feature-enhancing vector fields. A promising approach is the use of optimal steerable filters to derive an appropriate edge enhancing vector field =-=[30]-=-. This method uses filters that are more directional than the and components of the conventional gradient operator to derive a noise-resilient field. The general form of edge-based image energy can be... |

35 | Unifying boundary and region-based information for geodesic activetracking
- Paragios, Deriche
- 1999
(Show Context)
Citation Context ...rgy leads to a good segmentation. (a) Initialization, (b) edge only ( =1), (c) region only ( =0), and (d) unified ( =0:5). where . This unification is similar is philosophy to the approaches in [11], =-=[15]-=-. However, our scheme is more natural and yields a simpler expression since it combines the two energies into a single region integral. The simplicity of the expression will lead to computational adva... |

32 | Unsupervised contour representation and estimation using b-splines and a minimum description length criterion
- Figueiredo, Leito, et al.
- 2000
(Show Context)
Citation Context ...chemes which use local image information (typically gradient information) [3], [6], [9], [10], [17], [22], and 2) region-based methods, which use global image features (e.g., statistical formulation) =-=[8]-=-, [10], [11], [19], [23]–[27]. Since the best choice of the energy depends on the specific application at hand, we try to unify these approaches into a single framework; we obtain a general algorithm ... |

21 |
Leitão, “Bayesian estimation of ventricular contours in angiographic images
- Figueiredo, N
- 1992
(Show Context)
Citation Context ...image information (typically gradient information) [3], [6], [9], [10], [17], [22], and 2) region-based methods, which use global image features (e.g., statistical formulation) [8], [10], [11], [19], =-=[23]-=-–[27]. Since the best choice of the energy depends on the specific application at hand, we try to unify these approaches into a single framework; we obtain a general algorithm which can be tuned easil... |

17 |
Eric Weisstein’s world of mathematics,” http://mathworld.wolfram.com
- Weisstein
(Show Context)
Citation Context ... described). We approximate (46) by a discrete sum over the parameter . 6 In our experiments, looping arise in about 10% of the cases. 7 For a plane curve, the tangential angle is defined by d = jdrj =-=[39]-=-. Note that our criterion can give a value , even if the curve is looping (when ), which implies that it is not completely foolproof. In principle, it is possible to detect these cases by splitting th... |

14 |
Active contour model with gradient directional information: directional snake
- Park, Schoepflin, et al.
(Show Context)
Citation Context ...ice. Most of the commonly used approaches fall into two broadly defined categories: 1) edge-based schemes which use local image information (typically gradient information) [3], [6], [9], [10], [17], =-=[22]-=-, and 2) region-based methods, which use global image features (e.g., statistical formulation) [8], [10], [11], [19], [23]–[27]. Since the best choice of the energy depends on the specific application... |

12 | A unifying approach and interface for spline-based snakes
- Jacob, Blu, et al.
- 2001
(Show Context)
Citation Context ...local image information (typically gradient information) [3], [6], [9], [10], [17], [22], and 2) region-based methods, which use global image features (e.g., statistical formulation) [8], [10], [11], =-=[19]-=-, [23]–[27]. Since the best choice of the energy depends on the specific application at hand, we try to unify these approaches into a single framework; we obtain a general algorithm which can be tuned... |

11 |
Intelligent interactive image outlining using spline snakes
- Flickner, Sawhney, et al.
- 1994
(Show Context)
Citation Context ...ctions increases. In general, however, representations using smooth basis functions require fewer parameters than point-based approaches and, thus, result in faster optimization algorithms [6], [10], =-=[17]-=-. Moreover, such curve models have inherent regularity and hence do not require extra constraints to ensure smoothness [9], [17]. Since both the above mentioned schemes represent the curve explicitly,... |

10 | A deformable model for human organ extraction
- Gao, Kosaka, et al.
- 1998
(Show Context)
Citation Context ... term. The popular curve representation schemes in the snake literature are as follows: 1) point-based snakes, where the curve is an ordered collection of discrete points (also termed as snaxels) [3]–=-=[5]-=-; 2) parametric snakes, where the curve is described continuously in a parametric form, using basis functions such as B-splines [6]–[9], Fourier exponentials [10], [11], etc.; Manuscript received June... |

10 | An Exact method for Computing the Area Moments of Wavelet and Spline Curves
- Jacob, Blu, et al.
(Show Context)
Citation Context ... We compute the length as a discrete approximation of the integral as Length (44) The area of the curve is obtained by Green’s theorem as , which when expanded gives Area (45) where is obtained as in =-=[38]-=-. Note that the area obtained by the above expression is signed; its sign is utilized to determine the direction (clockwise or anticlockwise) of the curve.1240 IEEE TRANSACTIONS ON IMAGE PROCESSING, ... |

8 | B-spline active contours for fast video segmentation
- Precioso, Barlaud
- 2001
(Show Context)
Citation Context ...e curve knots to move tangential to the curve, thus bringing it to the curvilinear abscissa. An example of the type of improvement that can be obtained in this way is shown in Fig. 5. Precioso et al. =-=[33]-=- proposed to reparametrize the curve to the constant arc-length representation after each step of the optimization algorithm to avoid the curves from looping. This scheme would yield the same results ... |

7 | Comments in sinc interpolation of discret periodic signals
- Candocia, Prince
- 1998
(Show Context)
Citation Context ...ion and is equivalent to the Fourier representation of a 2 Scaling functions are functions that satisfy a two-scale relation '(x=2) = a(k)'(x 0 k), where a(k) is the two-scale mask [28]. closed curve =-=[29]-=-. The scaling function representation of a curve is given by where is the coefficient vector; they are often termed as knot points. We illustrate the parametric description of a polygon in terms of li... |

5 |
Simeltaneous inward and outward curve evolution
- Amadieu, Debreuve, et al.
- 1999
(Show Context)
Citation Context ....ch; michael.unser@epfl.ch). Digital Object Identifier 10.1109/TIP.2004.832919 3) geometric snakes, where the planar curve is represented as a level set of an appropriate two-dimensional surface [12]–=-=[16]-=-. The point-based approach can be viewed as a special case of parametric curve representation where the basis functions are uniform translates of a B-spline of degree zero; 1 likewise, parametric appr... |

4 | An active contour model for segmentation based on cubic B-splines and gradient vector flow - Gebhard, Mattes, et al. - 2001 |

3 |
Boundary fitting with parametrically deformable models
- Staib, Duncan
- 1992
(Show Context)
Citation Context ...mplexity is therefore proportional to . 2) Partial derivatives of the unified image energy. For closed curves, we preferentially use the unified energy to optimize the curve. In line with the work of =-=[10]-=-, [11], [37], we now use Green’s Theorem (12) to convert region integrals (over the region bounded by a closed curve) to integrals over the curve; our main motivation is computational efficiency. (20)... |

1 |
A 3-step algorithm using region based active contours for video objects detection
- Jehan-Besson, Barlaud, et al.
- 2002
(Show Context)
Citation Context ... information (typically gradient information) [3], [6], [9], [10], [17], [22], and 2) region-based methods, which use global image features (e.g., statistical formulation) [8], [10], [11], [19], [23]–=-=[27]-=-. Since the best choice of the energy depends on the specific application at hand, we try to unify these approaches into a single framework; we obtain a general algorithm which can be tuned easily to ... |