The structure of images

By Jan J. Koenderink. 2004

2,125 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Media Technology

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

In Media Technology

Citations in the corpus

Beyond Gaussian Pyramid: Multi-skip Feature Stacking for action recognition (2014, background): “quirement that no new detail information should be artificially found at the coarse scale of resolution [10].”
Multiscale Vision Transformers (2021, methodology): “In a parallel development, the computer vision community developed multiscale processing, sometimes called “pyramid” strategies, with Rosenfeld and Thurston [85], Burt and Adelson [8], Koenderink [61], among the key papers.”
Solving Linear Inverse Problems Using the Prior Implicit in a Denoiser (2020, background): “Moreover, the family of observation densities over different noise variances, p σ ( y ) , forms a Gaussian scale-space representation of the prior [28, 29], analogous to the temporal evolution of a diffusion process.”
Linear scale-space (1994, methodology): “Inner scale , outer scale , and scale - space Koenderink ( 1984 ) emphasized that the problem of scale must be faced in any imaging situation .”
BRIEF: Computing a Local Binary Descriptor Very Fast (2012, background): “The typical shape of the SURF-64 curve is a known artifact arising from approximating the Gaussian for scale-space analysis, degrading the performance under in-plane rotation for odd multiples of π/4 [32], [33].”
KAZE Features (2012, background): “Several authors [4,5] have shown that under some general assumptions, the Gaussian kernel and its set of partial derivatives are possible smoothing kernels for scale space analysis.”
Graph spectral image smoothing using the heat kernel (2008, methodology): “Thus, a linear isotropic diffusion at time t is equivalent to Gaussian filtering with spatial width √ 2t [4,5].”
A comparison of three image-object methods for the multiscale analysis of landscape structure (2003, background): “In particular, SS operators closely resemble receptive field profiles registered in mammalian retina and the visual cortex (Koenderink, 1984).”
Scaling Vision Transformers to Gigapixel Images via Hierarchical Self-Supervised Learning (2022, background): “Motivated by multiscale, pyramid-based image processing [11,42,61], recent progress in ViT architecture development has focused on efficiency and integration of multiscale information (e.”
Filling holes in complex surfaces using volumetric diffusion (2002, background): “In the terminology of image processing, these propagation and adding steps have been shown to be equivalent to spatial convolution and image compositing, respectively [11].”
Nonlinear Extraction of Independent Components of Natural Images Using Radial Gaussianization (2009, methodology): “On the other hand, the computer vision and image processing communities (partly motivated by the biological observations and partly by a desire to capture image features such as object boundaries) advocated the use of banks of oriented bandpass filters for representation and analysis of image data (”
Scale-Equivariant Steerable Networks (2019, background): “The differential structure of the image Koenderink (1984) allows one to make a connection between image formation mechanisms and the space of solutions of the 2-dimensional heat equation, which signiﬁcantly improved the image analysis models in the pre-deep learning era.”
3D Object Recognition in Cluttered Scenes with Local Surface Features: A Survey (2014, background): “It implies that any structure (feature) in a coarse scale should be able to find its cause in the finer scales [80].”
Speeded-Up Robust Features (SURF) (2008, background): “Gaussians are optimal for scale-space analysis [19,20], but in practice they have to be discretised and cropped (ﬁg-ure 2 left half).”
Affine Invariant Flows in the Beltrami Framework (2004, background): “The issue was first addressed in the context of linear scale-space [11, 24] where the invariance to Euclidean transformations is invoked.”
Stochastic Solutions for Linear Inverse Problems using the Prior Implicit in a Denoiser (2021, background): “Moreover, the family of observation densities over different noise variances, p � ( y ) , forms a Gaussian scale-space representation of the prior [33, 34], analogous to the temporal evolution of a diffusion process.”