A Survey of Gaussian Convolution Algorithms

  title={A Survey of Gaussian Convolution Algorithms},
  author={Pascal Getreuer},
  journal={Image Process. Line},
  • Pascal Getreuer
  • Published 17 December 2013
  • Computer Science
  • Image Process. Line
Gaussian convolution is a common operation and building block for algorithms in signal and image processing. Consequently, its ecient computation is important, and many fast approximations have been proposed. In this survey, we discuss approximate Gaussian convolution based on nite impulse response lters, DFT and DCT based convolution, box lters, and several recursive lters. Since boundary handling is sometimes overlooked in the original works, we pay particular attention to develop it here. We… 
Fast L1 Gaussian convolution via domain splitting
A fast and accurate approximation algorithm to convolve a L1 Gaussian function with images based on splitting a pixel domain into representative regions where it can efficiently perform discrete convolutions is proposed.
Computing an Exact Gaussian Scale-Space
This work discusses and numerically analyze the precision of three different alternatives for defining a discrete counterpart to the continuous Gaussian smoothing operator in the Gaussian scale-space, a family of increasingly blurred images.
Fast and Accurate Gaussian Pyramid Construction by Extended Box Filtering
A novel approach by investigating the extended box filters for an efficient Gaussian approximation is proposed, taking advantages of the cascade configuration, tiny kernels and memory cache, to develop a fast and suitable algorithm for embedded systems, typically smartphones.
OF IEEE ICIP ’ 14 Fast L 1 Gaussian Convolution via Domain Splitting
A fast and accurate approximation algorithm to convolve a L Gaussian function with images by splitting a pixel domain into representative regions where it can efficiently perform discrete convolutions.
Fast L 1 Gauss Transforms for Edge-Aware Image Filtering
A novel approximation method for fast Gaussian convolution of two-dimensional uniform point sets, such as 2D images that employs L1 distance metric for Gaussian function and domain splitting approach to achieve fast computation while preserving high accuracy.
Derivation and Analysis of Fast Bilinear Algorithms for Convolution
These algorithms can be used for convolution problems with multidimensional inputs or for filters larger than size of four, extending the state-of-the-art in Winograd-based convolution algorithms.
Fast Kernel Smoothing by a Low-Rank Approximation of the Kernel Toeplitz Matrix
The distribution of its eigenvalues is studied and it is shown that the upper bounds of the eigen values can be expressed analytically in terms of the Fourier transform of the kernel function.
An Efficient Description with Halide for IIR Gaussian Filter
RecFilter is extended to suit the methods commonly used in recursive Gaussian filter to solve the problem of approximation accuracy, and the proposed generator produces more computationally efficient codes and the resulting images have higher approximation accuracy.
Speed Up Bilateral Filtering via Sparse Approximation on a Learned Cosine Dictionary
This approach is the first method that adaptively constructs a cosine dictionary according to the input kernel and guarantees the best filtering accuracy and efficiency.
Rock image denoising by fast bilateral filtering
A fast denoising algorithm for 3D μCT images is proposed. The algorithm is based on an efficient implementation of bilateral filtering, in which the filter kernel is decomposed into a series of


Efficient and accurate Gaussian image filtering using running sums
  • Elhanan Elboher, M. Werman
  • Computer Science
    2012 12th International Conference on Intelligent Systems Design and Applications (ISDA)
  • 2012
A quadratic form kernel error function is proposed so that the SSD error of the output image is minimized and the proposed approach to approximate the Gaussian kernel by linear combination of constant functions results in a very efficient Gaussian filtering method.
Theoretical Foundations of Gaussian Convolution by Extended Box Filtering
Gaussian convolution is of fundamental importance in linear scale-space theory and in numerous applications. We introduce iterated extended box filtering as an efficient and highly accurate way to
Recursive implementation of the Gaussian filter
Efficient Synthesis of Gaussian Filters by Cascaded Uniform Filters
  • W. Wells
  • Computer Science
    IEEE Transactions on Pattern Analysis and Machine Intelligence
  • 1986
This paper describes an approach to implementing a Gaussian Pyramid which requires approximately two addition operations per pixel, per level, per dimension, and examines tradeoffs in choosing an algorithm for Gaussian filtering.
Stacked Integral Image
This work proposes a simplification for non-uniform filtering by stacking of box filters from a single integral image, and shows speedups of as much as 40∶1, similar to run time performance gains in kernel integral method, when comparing to naïve non uniform filtering.
Recursive Gaussian derivative filters
A strategy to design recursive implementations of the Gaussian filter and Gaussian regularized derivative filters that yield a high accuracy and excellent isotropy in n-D space is proposed.
Lattice architectures for multiple-scale gaussian convolution, image processing, sinusoid-based transforms and gabor filtering
A novel lattice architecture which performs multiple-scale Gaussian convolution of signals of any desired dimension based on the central limit theorem and involves repetitive convolution with small kernels to generate Gaussian smoothing.
Symmetric convolution and the discrete sine and cosine transforms
  • S. Martucci
  • Computer Science
    IEEE Trans. Signal Process.
  • 1994
The author defines symmetric convolution, relates the DSTs and DCTs to symmetric-periodic sequences, and then uses these principles to develop simple but powerful convolution-multiplication properties for the entire family of DST sine and cosine transforms.
Accurate and Efficient Computation of Gabor Features in Real-Time Applications
A novel framework in computation of Gabor features by utilizing a sophisticated interpolation scheme --- quadratic spline --- without increasing the overall computational complexity of the process is presented.
Fast Space-Variant Elliptical Filtering Using Box Splines
It is demonstrated that it is possible to filter an image with a Gaussian-like elliptic window of varying size, elongation and orientation using a fixed number of computations per pixel.