## Recognition Using Region Correspondences (1995)

### Cached

### Download Links

- [ftp.wisdom.weizmann.ac.il]
- [www.wisdom.weizmann.ac.il]
- [www.neci.nj.nec.com]
- DBLP

### Other Repositories/Bibliography

Venue: | International Journal of Computer Vision |

Citations: | 34 - 7 self |

### BibTeX

@ARTICLE{Basri95recognitionusing,

author = {Ronen Basri and David Jacobs},

title = {Recognition Using Region Correspondences},

journal = {International Journal of Computer Vision},

year = {1995},

volume = {25},

pages = {8--13}

}

### Years of Citing Articles

### OpenURL

### Abstract

A central problem in object recognition is to determine the transformation that relates the model to the image, given some partial correspondence between the two. This is useful in determining whether an object is present in an image, and if so, determining where the object is. We present a novel method of solving this problem that uses region information. In our approach the model is divided into volumes, and the image is divided into regions. Given a match between subsets of volumes and regions (without any explicit correspondence between different pieces of the regions) the alignment transformation is computed. The method applies to planar objects under similarity, affine, and projective transformations and to projections of 3-D objects undergoing affine and projective transformations. 1 Introduction A fundamental problem in recognition is pose estimation. Given a correspondence between some portions of an object model and some portions of an image, determine the transformation th...

### Citations

2510 |
Random sample consensus: A paradigm for model fitting with applications to image analysis and automated cartography
- Fischler, Bolles
- 1981
(Show Context)
Citation Context ...inguish shapes and determine pose using properties of an entire shape such as its moments, and local methods, which match local model and image features and use these matches to determine pose (e.g., =-=[10, 15, 22, 1, 19, 3]-=-). Global methods have the advantage of efficiency, since processing the image can be carried out independently of the model. MODEL IMAGE Figure 1: An example of region-based pose determination. The t... |

902 | Shape and motion from image streams under orthography: a factorization method
- Tpmasi, Kanade
- 1992
(Show Context)
Citation Context ... ~ t = (t x ; t y ; t z ). Then: u = t 11 x + t 12 y + t 13 z + t x v = t 21 x + t 22 y + t 23 z + t y (27) This projection model and its equivalent has been recently used by a number of researchers (=-=[29, 42, 26, 41, 22]-=-). It is also equivalent to applying scaled orthographic projection followed by a 2-D affine transformation ([22]), that is, taking a picture of a picture. Alternately, it is equivalent to a parapersp... |

846 |
Visual Pattern Recognition by Moment invariants
- Hu
- 1962
(Show Context)
Citation Context ...uniquely and present a number of experiments. The novelty of our method lies in its use of region information to determine pose. Current approaches to pose determination include global methods (e.g., =-=[12, 21]-=-) that distinguish shapes and determine pose using properties of an entire shape such as its moments, and local methods, which match local model and image features and use these matches to determine p... |

490 | Comparing images using the Hausdorff distance
- Huttenlocher, Klanderman, et al.
- 1993
(Show Context)
Citation Context ...aints is mapped to a distance of no more than jj from its target regions. (jj in this case represents a maximum norm, and so it is related to the Hausdorff metric. For work on Hausdorff matching, see =-=[13, 14, 2]-=-). Solving the system (10) may result in overcontraction. Consider, for example, the case of matching a single volume V to a single region R. Many transformations will shrink V to a small size and pla... |

448 |
Perceptual Organization and Visual Recognition
- Lowe
- 1984
(Show Context)
Citation Context ...inguish shapes and determine pose using properties of an entire shape such as its moments, and local methods, which match local model and image features and use these matches to determine pose (e.g., =-=[10, 15, 22, 1, 19, 3]-=-). Global methods have the advantage of efficiency, since processing the image can be carried out independently of the model. MODEL IMAGE Figure 1: An example of region-based pose determination. The t... |

412 |
Recognition by linear combinations of models
- Ullman, Basri
- 1991
(Show Context)
Citation Context ...inguish shapes and determine pose using properties of an entire shape such as its moments, and local methods, which match local model and image features and use these matches to determine pose (e.g., =-=[10, 15, 22, 1, 19, 3]-=-). Global methods have the advantage of efficiency, since processing the image can be carried out independently of the model. MODEL IMAGE Figure 1: An example of region-based pose determination. The t... |

328 |
Geometric hashing: A general and efficient model-based recognition scheme
- Lamdan, Wolfson
- 1988
(Show Context)
Citation Context ... ~ t = (t x ; t y ; t z ). Then: u = t 11 x + t 12 y + t 13 z + t x v = t 21 x + t 22 y + t 23 z + t y (27) This projection model and its equivalent has been recently used by a number of researchers (=-=[29, 42, 26, 41, 22]-=-). It is also equivalent to applying scaled orthographic projection followed by a 2-D affine transformation ([22]), that is, taking a picture of a picture. Alternately, it is equivalent to a parapersp... |

294 |
Affine structure from motion
- KOENDERINK, DOORN
- 1991
(Show Context)
Citation Context ... ~ t = (t x ; t y ; t z ). Then: u = t 11 x + t 12 y + t 13 z + t x v = t 21 x + t 22 y + t 23 z + t y (27) This projection model and its equivalent has been recently used by a number of researchers (=-=[29, 42, 26, 41, 22]-=-). It is also equivalent to applying scaled orthographic projection followed by a 2-D affine transformation ([22]), that is, taking a picture of a picture. Alternately, it is equivalent to a parapersp... |

255 |
Recognizing solid objects by alignment with an image
- Huttenlocher, Ullman
- 1990
(Show Context)
Citation Context |

250 |
A course in functional analysis
- Conway
- 1990
(Show Context)
Citation Context ...V 1 and T (V 2 ) ` V 2 . Since V 1 and V 2 are both closed and convex, and since T is a continuous transformation mapping these two volumes to inside themselves then, by Brouwer's fixed point theorem =-=[9]-=-, there exist two points ~p 1 2 V 1 and ~ p 2 2 V 2 that are fixed with respect to T , that is, T (~p i ) = ~ p i (i=1,2). (Note that ~ p 1 6= ~ p 2 since V 1 and V 2 are distinct.) Two points determi... |

248 | A paraperspective factorization method for shape and motion recovery
- Poelman, Kanade
- 1997
(Show Context)
Citation Context ...ng a picture of a picture. Alternately, it is equivalent to a paraperspective projection followed by translation ([6]), where paraperspective is a first-order approximation to perspective projection (=-=[33, 39]-=-). The forward constraint for the orthographic case becomes A(t 11 x + t 12 y + t 13 z + t x ) + B(t 21 x + t 22 y + t 23 z + t y ) + Cs0 (28) This constraint is linear in the transformation parameter... |

160 |
Pattern Classification and Scene Analysis. A Wileyinterscience publication
- Duda, Hart
- 1973
(Show Context)
Citation Context ...~c +s~ 1 (19) A solution to (18) exists if and only if a solution to (19) withs0 exists. (Note that other objective functions, e.g., the perceptron function, can be used for recovering ~ w, see e.g., =-=[13]-=- for a discussion of solutions to the linear discriminant functions problem.) Whens0 its value represents the minimal distance of a point to any line bounding the region (Figure 3). Maximizingsamounts... |

145 |
Finding edges and lines in image
- Canny
- 1983
(Show Context)
Citation Context ...ected by a plane always lead to a unique solution. 5 Experiments To test the scheme we took pictures of a number of roughly planar objects. We first processed these images using Canny's edge detector =-=[7]-=-. We then constructed polygonal approximations to the edges using Pavlidis and Horowitz's [20] split-and-merge algorithm. The resulting polygons approximate the original edges to within two pixels. Th... |

134 |
Invariant Descriptors for 3-D Object Recognition and Pose
- Forsyth, Mundy, et al.
- 1991
(Show Context)
Citation Context ...putationally expensive, requiring extensive search to find correspondences. Techniques have been applied to the recognition of curved objects that use algebraic descriptions of extended curves (e.g., =-=[18, 23, 11, 17]-=-) This approach can potentially overcome some of the difficulties of isolating local features reliably. However, the need for extended contour segments can make this approach vulnerable to partial occ... |

124 | A new approach for the recognition and positioning of two-dimensional objects
- Ayache, Faugeras
- 1986
(Show Context)
Citation Context |

119 | Affine invariant model-based object recognition - Lamdan, Schwartz, et al. - 1990 |

99 |
Three-dimensional model matching from an unconstrained viewpoint
- Thompson, Mundy
- 1987
(Show Context)
Citation Context ... to subsets of image features, and this match is used to recover the alignment transformation. This has been done using point features ([15, 17, 21, 42, 1, 2]), line segments ([30, 4, 36]), vertices (=-=[40]-=-), and distinguished points on curves such as inflection points or bitangents ([24, 37]). By relying on local properties, these methods can be more robust than global ones. Typically we must isolate a... |

96 |
Aircraft identification by moment invariants
- Dudani, Breeding, et al.
- 1977
(Show Context)
Citation Context ...ies of the object in the image are then computed and used to bring the object into the canonical description. Higher order moments, or other global descriptors may also be used to identify the object =-=[18, 35, 14, 32, 38, 34]-=-. The advantage of this approach is that it is computationally efficient, since processing the image can be carried out independently of the model. The main difficulty with this approach is that it re... |

93 |
Tracking Non-Rigid Objects in Complex Scenes
- Huttenlocher, Noh, et al.
- 1993
(Show Context)
Citation Context ...aints is mapped to a distance of no more than jj from its target regions. (jj in this case represents a maximum norm, and so it is related to the Hausdorff metric. For work on Hausdorff matching, see =-=[13, 14, 2]-=-). Solving the system (10) may result in overcontraction. Consider, for example, the case of matching a single volume V to a single region R. Many transformations will shrink V to a small size and pla... |

91 |
On Recognizing and Positioning Curved 3D objects from Image Contours
- Kriegman, Ponce
- 1990
(Show Context)
Citation Context ...putationally expensive, requiring extensive search to find correspondences. Techniques have been applied to the recognition of curved objects that use algebraic descriptions of extended curves (e.g., =-=[18, 23, 11, 17]-=-) This approach can potentially overcome some of the difficulties of isolating local features reliably. However, the need for extended contour segments can make this approach vulnerable to partial occ... |

88 |
Segmentation of plane curves
- Pavlidis, Horowitz
(Show Context)
Citation Context ...ictures of a number of roughly planar objects. We first processed these images using Canny's edge detector [7]. We then constructed polygonal approximations to the edges using Pavlidis and Horowitz's =-=[20]-=- split-and-merge algorithm. The resulting polygons approximate the original edges to within two pixels. Then, we extracted the roughly convex structures using Jacobs's grouping system [16]. The matchi... |

60 |
Three-Dimensional Moment Invariants
- Sadjadi, Hall
- 1980
(Show Context)
Citation Context ...ies of the object in the image are then computed and used to bring the object into the canonical description. Higher order moments, or other global descriptors may also be used to identify the object =-=[18, 35, 14, 32, 38, 34]-=-. The advantage of this approach is that it is computationally efficient, since processing the image can be carried out independently of the model. The main difficulty with this approach is that it re... |

59 |
Model-based Image Matching using Location
- Baird
- 1985
(Show Context)
Citation Context ...rom Equations (3) and (4) we obtain ~g T ~ ws\GammaC: (5) Baird pointed out that linear bounds on the location of a transformed model point leads to linear constraints on the feasible transformations =-=[4]-=- (see also [8, 6]). Affine transformation. If T is an affine transformation,sT is given in the form ` u v ' = ` a b c d '` x y ' + ` e f ' : (6) We can again rewrite the forward constraint as in Eq. (... |

59 | Canonical frames for planar object recognition
- Rothwell, Zisserman, et al.
- 1992
(Show Context)
Citation Context ...sformation. This has been done using point features ([15, 17, 21, 42, 1, 2]), line segments ([30, 4, 36]), vertices ([40]), and distinguished points on curves such as inflection points or bitangents (=-=[24, 37]-=-). By relying on local properties, these methods can be more robust than global ones. Typically we must isolate an entire shape to extract its global properties. However, we can often find local featu... |

49 |
Projective invariants of shapes
- Weiss
- 1988
(Show Context)
Citation Context ...putationally expensive, requiring extensive search to find correspondences. Techniques have been applied to the recognition of curved objects that use algebraic descriptions of extended curves (e.g., =-=[18, 23, 11, 17]-=-) This approach can potentially overcome some of the difficulties of isolating local features reliably. However, the need for extended contour segments can make this approach vulnerable to partial occ... |

45 | The alignment of objects with smooth surfaces - Basri, Ullman - 1993 |

34 |
Three-Dimensional Shape Analysis Using Moments and Fourier Descriptors
- Reeves, Prokop, et al.
- 1988
(Show Context)
Citation Context ...ies of the object in the image are then computed and used to bring the object into the canonical description. Higher order moments, or other global descriptors may also be used to identify the object =-=[18, 35, 14, 32, 38, 34]-=-. The advantage of this approach is that it is computationally efficient, since processing the image can be carried out independently of the model. The main difficulty with this approach is that it re... |

32 |
New methods for matching 3D objects with single perspective view
- Horaud
- 1987
(Show Context)
Citation Context ... model and from the image. Subsets of model features are matched to subsets of image features, and this match is used to recover the alignment transformation. This has been done using point features (=-=[15, 17, 21, 42, 1, 2]-=-), line segments ([30, 4, 36]), vertices ([40]), and distinguished points on curves such as inflection points or bitangents ([24, 37]). By relying on local properties, these methods can be more robust... |

32 | Efficient Model Library Access by Projectively Invariant Indexing Functions
- Rothwell, Zisserman, et al.
- 1992
(Show Context)
Citation Context ...del features are matched to subsets of image features, and this match is used to recover the alignment transformation. This has been done using point features ([15, 17, 21, 42, 1, 2]), line segments (=-=[30, 4, 36]-=-), vertices ([40]), and distinguished points on curves such as inflection points or bitangents ([24, 37]). By relying on local properties, these methods can be more robust than global ones. Typically ... |

28 | Space efficient 3D model indexing
- Jacobs
- 1992
(Show Context)
Citation Context |

20 |
Bounded boxes, Hausdorff distance, and a new proof of an interesting Helly-type theorem
- Amenta
- 1994
(Show Context)
Citation Context ...aints is mapped to a distance of no more than jj from its target regions. (jj in this case represents a maximum norm, and so it is related to the Hausdorff metric. For work on Hausdorff matching, see =-=[13, 14, 2]-=-). Solving the system (10) may result in overcontraction. Consider, for example, the case of matching a single volume V to a single region R. Many transformations will shrink V to a small size and pla... |

18 |
Recognizing 3-D Objects Using 2D Images
- Jacobs
- 1992
(Show Context)
Citation Context ... Horowitz's [20] split-and-merge algorithm. The resulting polygons approximate the original edges to within two pixels. Then, we extracted the roughly convex structures using Jacobs's grouping system =-=[16]-=-. The matching between the regions was specified manually. Finally, the transformations relating these images were recovered using either the forward or backward solutions. Figure 6 shows an image of ... |

17 | Hot curves for modelling and recognition of smooth curved 3d shapes
- Joshi, Ponce, et al.
- 1994
(Show Context)
Citation Context ...sformation. This has been done using point features ([15, 17, 21, 42, 1, 2]), line segments ([30, 4, 36]), vertices ([40]), and distinguished points on curves such as inflection points or bitangents (=-=[24, 37]-=-). By relying on local properties, these methods can be more robust than global ones. Typically we must isolate an entire shape to extract its global properties. However, we can often find local featu... |

16 |
The real projective plane
- Coxeter
- 1993
(Show Context)
Citation Context ... 0; 1) T on l. The cross ratio of the four points is given by C(ff; fi) = ff(fi \Gamma 1) fi(ff \Gamma 1) : A collineation of the plane preserves the cross ratio. (See a standard text, such as Coxeter=-=[12]-=- for further discussion of the cross ratio). We can therefore write C(T (ff); T (fi)) = C(ff; fi): 16 23 3 p p p 1 2 l 12 l l 13 Figure 6: Motion of points between three fixed points. The motion is a-... |

15 |
Tracking Non rigid Objects
- Huttenlocher, Noh, et al.
- 1993
(Show Context)
Citation Context ...aints is mapped to a distance of no more than jj from its target regions. (jj in this case represents a maximum norm, and so it is related to the Hausdorff metric. For work on Hausdorff matching, see =-=[19, 20]-=-. Also, [3] specifically discusses the efficient Hausdorff matching of convex shapes undergoing translation and scaling.) Solving the system (19) may result in over-contraction. Consider, for example,... |

14 |
Fast and robust 3D recognition by alignment
- Alter, Grimson
- 1993
(Show Context)
Citation Context ... model and from the image. Subsets of model features are matched to subsets of image features, and this match is used to recover the alignment transformation. This has been done using point features (=-=[15, 17, 21, 42, 1, 2]-=-), line segments ([30, 4, 36]), vertices ([40]), and distinguished points on curves such as inflection points or bitangents ([24, 37]). By relying on local properties, these methods can be more robust... |

12 |
Identification of threedimensional objects using Fourier descriptors of the boundary curve
- Richard, Hemami
- 1974
(Show Context)
Citation Context ...uniquely and present a number of experiments. The novelty of our method lies in its use of region information to determine pose. Current approaches to pose determination include global methods (e.g., =-=[12, 21]-=-) that distinguish shapes and determine pose using properties of an entire shape such as its moments, and local methods, which match local model and image features and use these matches to determine p... |

11 |
Error propagation in full 3d from 2d object recognition
- Alter, Jacobs
- 1994
(Show Context)
Citation Context |

10 | Model based recognition using pruned correspondence search - Breuel - 1991 |

10 |
Perceptual organization and visual recognition,” The Netherlands
- Lowe
- 1985
(Show Context)
Citation Context ...del features are matched to subsets of image features, and this match is used to recover the alignment transformation. This has been done using point features ([15, 17, 21, 42, 1, 2]), line segments (=-=[30, 4, 36]-=-), vertices ([40]), and distinguished points on curves such as inflection points or bitangents ([24, 37]). By relying on local properties, these methods can be more robust than global ones. Typically ... |

6 |
Polynomial Time Object Recognition in the Presence of Clutter, Occlusion and Uncertainty
- Cass
- 1992
(Show Context)
Citation Context ...(3) and (4) we obtain ~g T ~ ws\GammaC: (5) Baird pointed out that linear bounds on the location of a transformed model point leads to linear constraints on the feasible transformations [4] (see also =-=[8, 6]-=-). Affine transformation. If T is an affine transformation,sT is given in the form ` u v ' = ` a b c d '` x y ' + ` e f ' : (6) We can again rewrite the forward constraint as in Eq. (5), letting ~ w T... |

5 |
3D object recognition by combination of perspective images
- Sugimoto, Murota
- 1993
(Show Context)
Citation Context ...ng a picture of a picture. Alternately, it is equivalent to a paraperspective projection followed by translation ([6]), where paraperspective is a first-order approximation to perspective projection (=-=[33, 39]-=-). The forward constraint for the orthographic case becomes A(t 11 x + t 12 y + t 13 z + t x ) + B(t 21 x + t 22 y + t 23 z + t y ) + Cs0 (28) This constraint is linear in the transformation parameter... |

4 |
Shape descimination using Fourier descriptors
- Persoon, Fu
- 1977
(Show Context)
Citation Context |

3 |
Recognition using region correspondences", Forthcoming
- Basri, Jacobs
- 1995
(Show Context)
Citation Context ...s, rather than local boundary features. In this paper we will focus primarily on developing this method for the recognition of a planar model in a 2-D image, taken from an arbitrary 3-D viewpoint. In =-=[5]-=- we show that the method can be extended to the recognition of a 3D object in a 2-D image. In summary, our method has the following advantages: region boundaries need not be precisely localized; the c... |

2 |
Projective invariants of shape", DARPA Image Unerstanding Workshop: 1125--1134
- Weiss
- 1988
(Show Context)
Citation Context ...t are intermediate between the local and global methods. These recognition methods match algebraic descriptions of the surfaces of objects to the curves found in images. This is done, for example, by =-=[27, 43, 16, 25]-=-. These methods rely on locating extended portions of an object's contour, but not the entire contour. This approach can potentially overcome some of the difficulties of isolating local features relia... |

1 |
Bounded and Unbounded
- Keren, Subrahmonia, et al.
- 1992
(Show Context)
Citation Context |

1 |
Paraperspective j affine", The Weizmann
- Basri
- 1994
(Show Context)
Citation Context ... orthographic projection followed by a 2-D affine transformation ([22]), that is, taking a picture of a picture. Alternately, it is equivalent to a paraperspective projection followed by translation (=-=[6]-=-), where paraperspective is a first-order approximation to perspective projection ([33, 39]). The forward constraint for the orthographic case becomes A(t 11 x + t 12 y + t 13 z + t x ) + B(t 21 x + t... |