## Indexing based on scale invariant interest points (2001)

### Cached

### Download Links

Venue: | In Proceedings of the 8th International Conference on Computer Vision |

Citations: | 335 - 28 self |

### BibTeX

@INPROCEEDINGS{Mikolajczyk01indexingbased,

author = {Krystian Mikolajczyk and Cordelia Schmid},

title = {Indexing based on scale invariant interest points},

booktitle = {In Proceedings of the 8th International Conference on Computer Vision},

year = {2001},

pages = {525--531}

}

### Years of Citing Articles

### OpenURL

### Abstract

This paper presents a new method for detecting scale invariant interest points. The method is based on two recent results on scale space: 1) Interest points can be adapted to scale and give repeatable results (geometrically stable). 2) Local extrema over scale of normalized derivatives indicate the presence of characteristic local structures. Our method first computes a multi-scale representation for the Harris interest point detector. We then select points at which a local measure (the Laplacian) is maximal over scales. This allows a selection of distinctive points for which the characteristic scale is known. These points are invariant to scale, rotation and translation as well as robust to illumination changes and limited changes of viewpoint. For indexing, the image is characterized by a set of scale invariant points; the scale associated with each point allows the computation of a scale invariant descriptor. Our descriptors are, in addition, invariant to image rotation, to affine illumination changes and robust to small perspective deformations. Experimental results for indexing show an excellent performance up to a scale factor of 4 for a database with more than 5000 images. 1

### Citations

1692 | Object recognition from local scale-invariant features
- Lowe
- 1999
(Show Context)
Citation Context ...should also be invariant to image transformations. Recent methods for indexing differ in the type of invariants used. Rotation invariants have been presented by [10], rotation and scale invariants by =-=[8]-=- and affine invariants by [13]. Schmid and Mohr [10] extract a set of interest points and characterize each of the points by rotationally invariThis work was supported by the French project RNRT AGIR.... |

876 | The design and use of steerable filters
- Freeman, Adelson
- 1991
(Show Context)
Citation Context ...Descriptors. Our descriptors are Gaussian derivatives which are computed at the characteristic scale. Invariance to rotation is obtained by "steering" the derivatives in the direction of the=-= gradient [4]-=-. To obtain a stable estimation of the gradient direction, we use the peak in a histogram of local gradient orientations. Invariance to the affine intensity changes is obtained by dividing the derivat... |

557 |
Scale-space filtering
- Witkin
- 1983
(Show Context)
Citation Context ...elect the characteristic scale. We then present experimental results for scale selection. Scale-space. The scale-space representation is a set of images represented at different levels of resolutions =-=[14]-=-. Different levels of resolution are in general created by convolution with the Gaussian kernel: L(x; s) = G(s)sI(x) with I the image and x = (x; y). We can represent a feature (i.e. edges, corners) a... |

513 | Feature detection with automatic scale selection
- Lindeberg
- 1998
(Show Context)
Citation Context ...scale space: 1) Interest points can be adapted to scale and give repeatable results [3]. 2) Local extrema over scale of normalized derivatives indicate the presence of characteristic local structures =-=[7]-=-. The first step of our approach is to compute interest points at several scale levels. We then select points at which a local measure (the Laplacian) is maximal over scales. This allows to select a s... |

469 | Local Grayvalue Invariants for Image Retrieval
- Schmid, Mohr
- 1997
(Show Context)
Citation Context ... to changes of viewing conditions they should also be invariant to image transformations. Recent methods for indexing differ in the type of invariants used. Rotation invariants have been presented by =-=[10]-=-, rotation and scale invariants by [8] and affine invariants by [13]. Schmid and Mohr [10] extract a set of interest points and characterize each of the points by rotationally invariThis work was supp... |

256 | Reliable feature matching across widely separated views
- Baumberg
(Show Context)
Citation Context ...descriptors at these scales. An object is represented by the set of these descriptors. All of the above methods are limited to a scale factor of 2. Similar approaches exist for wide-baseline matching =-=[1, 3, 5, 9, 12]-=-. The problem is however more restricted. Additional constraints can be imposed and the search complexity is less prohibitive. For example, Prichett and Zisserman [9] first match regions bound by four... |

138 | Wide baseline stereo matching
- Pritchett, Zisserman
- 1998
(Show Context)
Citation Context ...descriptors at these scales. An object is represented by the set of these descriptors. All of the above methods are limited to a scale factor of 2. Similar approaches exist for wide-baseline matching =-=[1, 3, 5, 9, 12]-=-. The problem is however more restricted. Additional constraints can be imposed and the search complexity is less prohibitive. For example, Prichett and Zisserman [9] first match regions bound by four... |

91 | Comparing and evaluating interest points
- Schmid, Mohr, et al.
- 1998
(Show Context)
Citation Context ...ector is used for 2D localization as it has shown to be most reliable in the presence of image rotation, illumination transformations and perspective deformations as shown in a comparative evaluation =-=[11]-=-. However, the repeatability of this detector fails when the resolution of images changes significantly. In order to deal with such changes, the Harris detector has to be adapted to the scale factor [... |

87 | Gool. ”Content-based image retrieval based on local affinity invariant regions
- Tuytelaars, Van
- 1999
(Show Context)
Citation Context ...mage transformations. Recent methods for indexing differ in the type of invariants used. Rotation invariants have been presented by [10], rotation and scale invariants by [8] and affine invariants by =-=[13]-=-. Schmid and Mohr [10] extract a set of interest points and characterize each of the points by rotationally invariThis work was supported by the French project RNRT AGIR. ant descriptors which are com... |

64 | Matching images with different resolutions
- Dufournaud, Schmid, et al.
- 2000
(Show Context)
Citation Context ...descriptors at these scales. An object is represented by the set of these descriptors. All of the above methods are limited to a scale factor of 2. Similar approaches exist for wide-baseline matching =-=[1, 3, 5, 9, 12]-=-. The problem is however more restricted. Additional constraints can be imposed and the search complexity is less prohibitive. For example, Prichett and Zisserman [9] first match regions bound by four... |

56 |
Geometric information criterion for model selection
- Kanatani
- 1998
(Show Context)
Citation Context ... on RANdom SAmple Consensus (RANSAC) allows to reject inconsistent matches. For our experimental results the transformation is either a homography or a fundamental matrix. A model selection algorithm =-=[6]-=- can of course be used to automatically decide what transformation is the most appropriate one. Indexing. A voting algorithm is used to select the most similar images in the database. This makes retri... |

55 |
Wide baseline point matching using affine invariants computed from intensity profiles
- Tell, Carlsson
- 2000
(Show Context)
Citation Context |

19 |
Local scale selection for gaussian based description techniques
- Chomat, Verdière, et al.
- 2000
(Show Context)
Citation Context ...ffine invariant regions and describing them by color invariants. To find these regions they simultaneously use interest points and contours. Instead of using an initial set of features, Chomat et al. =-=[2]-=- select the appropriate scale for every point in the image and compute descriptors at these scales. An object is represented by the set of these descriptors. All of the above methods are limited to a ... |

8 | Multiscale image registration using scale trace correlation
- Hansen, Morse
- 1999
(Show Context)
Citation Context |