Scale & Affine Invariant Interest Point Detectors

By Krystian Mikolajczyk, Cordelia Schmid. 2004

Subfields: Computer Vision and Pattern Recognition

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

Soft-NMS — Improving Object Detection with One Line of Code (2017, methodology): “Subsequently, it has been applied to multiple tasks like feature point detection [19, 12, 20], face detection [29] and object detection [4, 8, 10].”
End-to-End Learning of Deep Visual Representations for Image Retrieval (2016, methodology): “We use the SIFT and Hessian-Affine keypoint detectors (Lowe 2004; Mikolajczyk and Schmid 2004) and match keypoints using the first-to-second neighbor ratio rule (Lowe 2004).”
Kornia: an Open Source Differentiable Computer Vision Library for PyTorch (2019, methodology): “The module provides differentiable versions of the Harris corner detector[20], Hessian detector [6], their scale and affine covariant versions [38], DoG [33], patch dominant gradient orientation [33] and the SIFT descriptor [33].”
Semantic texton forests for image categorization and segmentation (2008, unknown): “[19]) or more densely; recent results in [22] suggest that densely sampling visual words improves categorization performance.”
Scene Classification Using a Hybrid Generative/Discriminative Approach (2008, background): “structed by elliptical shape adaptation about an interest point [ 22 ].”
Measuring the Objectness of Image Windows (2011, methodology): “Interest point detectors (IPs) [29], [41] respond to local textured image neighborhoods and are widely used for finding image correspondences [41] and recognizing specific objects [37].”
Merging chrominance and luminance in early, medium, and late fusion using Convolutional Neural Networks (2015, methodology): “Similarly, the Harris-Laplace detector [Mikolajczyk and Schmid, 2004; Mikolajczyk et al., 2005a] detects also scale and affine-invariant regions.”
Dense interest points (2010, methodology): “We compare our dense interest points with patches extracted by dense sampling on the same regular grid, and with the Hessian-Laplace scaleinvariant interest point detector of [ 14 ].”
Linear Time Maximally Stable Extremal Regions (2008, background): “Maximally Stable Extremal Regions (MSER) described by Matas el al [3] have become one of the commonly used region detector types, partly because of their high repeatability and partly because they are somewhat complementary to many other commonly used detectors [7, 4 ,13].”
Image Matching from Handcrafted to Deep Features: A Survey (2020, methodology): “Another classical blob feature detection strategy is based on the determinant of Hessian (DoH) (Mikolajczyk and Schmid 2001, 2004).”
R2D2: Repeatable and Reliable Detector and Descriptor (2019, methodology): “Following [27], we compute the repeatability score for a pair of images as the number of point correspondences found between the two images divided by the minimum number of keypoint detections in the image pair.”
A 2-D wavelet decomposition-based bag-of-visual-words model for land-use scene classification (2014, background): “Generally, it samples an image efficiently with various local interest point detectors (Lowe 2004; Mikolajczyk and Schmid 2004) or dense regions (Lazebnik, Schmid, and Ponce 2006; Qin and Yung 2010), and describes it with local descriptors (Mikolajczyk and Schmid 2005) among which affine invariant…”
Harris 3D: a robust extension of the Harris operator for interest point detection on 3D meshes (2011, methodology): “Second, Harris-based methods have been effectively used in a number of applications and they have a high effectiveness as reported in the evaluation reports [19, 21].”
Robust supervised classification with mixture models: Learning from data with uncertain labels (2009, methodology): “The local image descriptors were obtained by first using the Harris-Laplace detector [22] to extract interest points and by then using the SIFT descriptor [20] to represent the scale-invariant regions around these points.”
Near Duplicate Image Detection: min-Hash and tf-idf Weighting (2008, background): “Popular choices are MSER [15], DoG (difference of Gaussians) [14] or multi-scale Hessian interest points [16].”