FREAK: Fast Retina Keypoint

By Alexandre Alahi, Raphael Ortiz, Pierre Vandergheynst. 2012 IEEE Conference on Computer Vision and Pattern Recognition, 2012

1,982 citations recorded by the source corpus.

A large number of vision applications rely on matching keypoints across images. The last decade featured an arms-race towards faster and more robust keypoints and association algorithms: Scale Invariant Feature Transform (SIFT)[17], Speed-up Robust Feature (SURF)[4], and more recently Binary Robust Invariant Scalable Keypoints (BRISK)[I6] to name a few. These days, the deployment of vision algorithms on smart phones and embedded devices with low memory and computation complexity has even upped the ante: the goal is to make descriptors faster to compute, more compact while remaining robust to scale, rotation and noise. To best address the current requirements, we propose a novel keypoint descriptor inspired by the human visual system and more precisely the retina, coined Fast Retina Keypoint (FREAK). A cascade of binary strings is computed by efficiently comparing image intensities over a retinal sampling pattern. Our experiments show that FREAKs are in general faster to compute with lower memory load and also more robust than SIFT, SURF or BRISK. They are thus competitive alternatives to existing keypoints in particular for embedded applications.

Subfields: Aerospace Engineering, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/bfe2e9b42cca691dcddca8ca90a0d75c67cead58

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Aerospace Engineering

In Computer Vision and Pattern Recognition

Citations in the corpus

Image Reconstruction from Bag-of-Visual-Words (2014, methodology): “[8] converted BREAF descriptors [2] and FREAK descriptors [1] into image patches.”
Multi-View Stereo: A Tutorial (2015, background): “The development of fast and high quality feature detectors [87, 135, 57] and descriptors [135, 36, 159, 130, 26] was a crucial development towards making SfM work with unstructured datasets.”
KETO: Learning Keypoint Representations for Tool Manipulation (2019, background): “Most keypoint representations are either defined as low-level features in detection and matching [32], [47], [2] or used to describe object parts on a category level [24], [56].”
Latent-Class Hough Forests for 3D Object Detection and Pose Estimation (2014, background): “While this approach has shown great results on highly textured objects [2, 5, 11, 48, 61], in the case of low-textured objects, unseen background clutter can provide significant false positive correspondences.”
Pose Estimation for Augmented Reality: A Hands-On Survey (2016, positive): “Pattern for BRIEF [21] (left), BRISK [69] (center), Freak [1] (right).”
Image Matching from Handcrafted to Deep Features: A Survey (2020, methodology): “Inspired by the retina structure, Alahi et al. (2012) proposed the FREAK descriptor by comparing image intensities over a retinal sampling pattern for fast computing andmatchingwith lowmemory costwhile remaining robust to scale, rotation, and noise.”
Towards Reversal-Invariant Image Representation (2017, background): “Other examples include Shape Context (Belongie et al. 2002), SURF (Bay et al. 2008),BRIEF (Calonder et al. 2010), BRISK (Leutenegger et al. 2011), ORB (Rublee et al. 2011), FREAK (Alahi et al. 2012), etc. Radial transform (Takacs et al. 2013) and polar analysis (Liu et al. 2014) play important roles”
LIFT: Learned Invariant Feature Transform (2016, methodology): “…detector [36], MROGH [37] with Harris-aﬃne detector, LIOP [38] with Harris-aﬃne detector, BiCE [39] with Edge Foci detector [19], BRISK [40], FREAK [41] with BRISK detector, VGG [23] (trained on ‘Liberty’) with SIFT detector; DeepDesc [10] with SIFT detector; PN-Net [25] with SIFT detector.”
Fast Exact Search in Hamming Space With Multi-Index Hashing (2013, positive): “Nearest neighbor (NN) search on binary codes is used for image search [29], [35], [38], matching local features [2], [7], [16], [33], image classiﬁcation [6], object segmentation [19], and parameter estimation [32].”
A review of 3D reconstruction techniques in civil engineering and their applications (2018, methodology): “Inputs Steps Algorithms and methods Monocular images/video frames1 Feature extraction SIFT [28,39–43] (including SIFT detector and SIFT descriptor); SURF [21] (including SURF detector and SURF descriptor); ASIFT [44]; FREAK [45]; Harris corner detector [38]; FAST [38]; MSD [38] Feature matching RANd”
Get Out of My Lab: Large-scale, Real-Time Visual-Inertial Localization (2015, background): “To enable real-time performance on mobile devices efficient binary descriptors [2] are used and projected to a real-valued space following [7, 29].”
Median Robust Extended Local Binary Pattern for Texture Classification (2016, methodology): “mation, inspired by DAISY [23], BRISK [24] and FREAK [25].”
A review of multimodal image matching: Methods and applications (2021, background): “Other binary selection and comparison rules include concentric circle sampling strategy with increasing radius [209] and comparing image intensities over a retinal sampling pattern [210].”
Learning Compact Binary Descriptors with Unsupervised Deep Neural Networks (2016, methodology): “Binary Descriptors: Earlier works related to binary descriptors can be traced back to BRIEF [6], ORB [33], BRISK [22], and FREAK [1].”
Made to measure: Bespoke landmarks for 24-hour, all-weather localisation with a camera (2016, methodology): “These are then described with a local feature descriptor such as SIFT [10], SURF [11] or one of the binary descriptors [12][13][14][15].”