Author

Pascal Getreuer

30 papers · 2006–2026

2006peak 8 in 20112026

Frequent co-authors

Papers

showing 1–25 of 30 RSS

Rudin-Osher-Fatemi Total Variation Denoising using Split Bregman

Computer Vision and Pattern Recognition · Image Processing On Line · 19 May 2012 · cited 265× · DOI ↗ · more like this ·

Denoising is the problem of removing noise from an image. The most commonly studied case is with additive white Gaussian noise (AWGN), where the observed noisy image f is related to the underlying true image u by f=u+η and η is at each point in space independently and identically distributed as a zero-mean Gaussian random variable. Total variation (TV) regularization is a technique that was originally developed for AWGN image denoising by Rudin, Osher, and Fatemi. The TV regularization technique has since been applied to a multitude of other imaging problems, see for example Chan and Shen's book. We focus here on the split Bregman algorithm of Goldstein and Osher for TV-regularized denoising.

Chan-Vese Segmentation

Pascal Getreuer

Computer Vision and Pattern Recognition · Image Processing On Line · 8 Aug 2012 · cited 201× · DOI ↗ · more like this ·

While many segmentation methods rely heavily in some way on edge detection, the "Active Contours Without Edges" method by Chan and Vese ignores edges completely. Instead, the method optimally fits a two-phase piecewise constant model to the given image. The segmentation boundary is represented implicitly with a level set function, which allows the segmentation to handle topological changes more easily than explicit snake methods. This article describes the level set formulation of the Chan–Vese model and its numerical solution using a semi-implicit gradient descent. We also discuss the Chan–Sandberg–Vese method, a straightforward extension of Chan–Vese for vector-valued images.

Trainable Frontend For Robust and Far-Field Keyword Spotting

Yuxuan Wang, Pascal Getreuer, Thad Hughes +2 more

cs.CL · 19 Jul 2016 · cited 165× · arXiv ↗ · more like this ·

Robust and far-field speech recognition is critical to enable true hands-free communication. In far-field conditions, signals are attenuated due to distance. To improve robustness to loudness variation, we introduce a novel frontend called per-channel energy normalization (PCEN). The key ingredient of PCEN is the use of an automatic gain control based dynamic compression to replace the widely used static (such as log or root) compression. We evaluate PCEN on the keyword spotting task. On our large rerecorded noisy and far-field eval sets, we show that PCEN significantly improves recognition performance. Furthermore, we model PCEN as neural network layers and optimize high-dimensional PCEN parameters jointly with the keyword spotting acoustic model. The trained PCEN frontend demonstrates significant further improvements without increasing model complexity or inference-time cost.

A Survey of Gaussian Convolution Algorithms

Pascal Getreuer

Computer Vision and Pattern Recognition · Image Processing On Line · 17 Dec 2013 · cited 126× · DOI ↗ · more like this ·

Gaussian convolution is a common operation and building block for algorithms in signal and image processing. Consequently, its efficient computation is important, and many fast approximations have been proposed. In this survey, we discuss approximate Gaussian convolution based on finite impulse response filters, DFT and DCT based convolution, box filters, and several recursive filters. Since boundary handling is sometimes overlooked in the original works, we pay particular attention to develop it here. We perform numerical experiments to compare the speed and quality of the algorithms.

Automatic Color Enhancement (ACE) and its Fast Implementation

Pascal Getreuer

Computer Vision and Pattern Recognition · Image Processing On Line · 6 Nov 2012 · cited 122× · DOI ↗ · more like this ·

Automatic Color Enhancement "ACE" is an effective method for color image enhancement introduced by Gatta, Rizzi, and Marini based on modeling several low level mechanisms of the human visual system. The direct computation of ACE on an NxN image costs O(N^4) operations. This article describes two fast approximations of ACE. First, the algorithm of Bertalmío, Caselles, Provenzi, and Rizzi uses a polynomial approximation of the slope function to decomposes the main computation into convolutions, reducing the cost to O(N^2 log N). Second, an algorithm based on interpolating intensity levels also reduces the main computation to convolutions. The use of ACE for image enhancement and color correction is demonstrated.

Linear Methods for Image Interpolation

Pascal Getreuer

Computer Vision and Pattern Recognition · Image Processing On Line · 27 Sep 2011 · cited 121× · DOI ↗ · more like this ·

We discuss linear methods for interpolation, including nearest neighbor, bilinear, bicubic, splines, and sinc interpolation. We focus on separable interpolation, so most of what is said applies to one-dimensional interpolation as well as N-dimensional separable interpolation.

Total Variation Inpainting using Split Bregman

Pascal Getreuer

Computer Vision and Pattern Recognition · Image Processing On Line · 30 Jul 2012 · cited 106× · DOI ↗ · more like this ·

Given an image where a specified region is unknown, image inpainting or image completion is the problem of inferring the image content in this region. Traditional retouching or inpainting is the practice of restoring aged artwork, where damaged or missing portions are repainted based on the surrounding content to approximate the original appearance. In the context of digital images, inpainting is used to restore regions of an image that are corrupted by noise or where the data is missing. Inpainting is also used to solve disocclusion, to estimate the scene behind an obscuring foreground object. A popular use of digital inpainting is object removal, for example, to remove a trashcan that disrupts a scene of otherwise natural beauty. Inpainting is an interpolation problem, filling the unknown region with a condition to agree with the known image on the boundary. A classical solution for…

Contour Stencils: Total Variation along Curves for Adaptive Image Interpolation

Pascal Getreuer

Computer Vision and Pattern Recognition · SIAM Journal on Imaging Sciences · 2011 · cited 39× · DOI ↗ · more like this ·

Image interpolation is the problem of increasing the resolution of a given image. An important aspect of interpolation is accurate estimation of edge orientations. This work introduces contour stencils, a new method for estimating the image contours based on total variation along curves. Contour stencils are able to distinguish lines of different orientations, curves, corners, and other geometric features with a computationally efficient formula. Contour stencils are applied in designing an edge directed color interpolation method. The method incorporates an efficient approximation of deconvolution. Although most interpolation methods that involve deconvolution require either solving a large linear system or running many iterations, this method has linear complexity in the number of pixels and can be computed in one or a small number of passes through the image. Comparisons show that…

Contour stencils for edge-adaptive image interpolation

Pascal Getreuer

Computer Vision and Pattern Recognition · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 8 Dec 2008 · cited 35× · DOI ↗ · more like this ·

We first develop a simple method for detecting the local orientation of image contours and then use this detection to design an edge-adaptive image interpolation strategy. The detection is based on total variation: small total variation along a candidate curve implies that the image is approximately constant along that curve, which suggests it is a good approximation to the contours. The proposed strategy is to measure the total variation over a "contour stencil," a set of parallel curves localized over a small patch in the image. This contour stencil detection is used to design an edge-adaptive image interpolation strategy. The interpolation is computationally efficient, operates robustly over a variety of image features, and performs competitively in a comparison against existing methods. The method extends readily to vector-valued data and is demonstrated for color image…

BLADE: Filter Learning for General Purpose Computational Photography

Pascal Getreuer, Ignacio Garcia-Dorado, John Isidoro +3 more

cs.CV · November 2017 · cited 29× · arXiv ↗ · more like this ·

The Rapid and Accurate Image Super Resolution (RAISR) method of Romano, Isidoro, and Milanfar is a computationally efficient image upscaling method using a trained set of filters. We describe a generalization of RAISR, which we name Best Linear Adaptive Enhancement (BLADE). This approach is a trainable edge-adaptive filtering framework that is general, simple, computationally efficient, and useful for a wide range of problems in computational photography. We show applications to operations which may appear in a camera pipeline including denoising, demosaicing, and stylization.

Image Demosaicking with Contour Stencils

Pascal Getreuer

Computer Vision and Pattern Recognition · Image Processing On Line · 24 Mar 2012 · cited 27× · DOI ↗ · more like this ·

Demosaicking (or demosaicing) is the problem of interpolating full color information on an image where only one color component is known at each pixel. Most demosaicking methods involve some kind of estimation of the underlying image structure, for example, choosing adaptively between interpolating in the horizontal or vertical direction. This article discusses the implementation details of the method introduced in Getreuer, “Color Demosaicing with Contour Stencils,” 2011. Mosaicked contour stencils first estimate the image contour orientations directly from the mosaicked data. The mosaicked contour stencils are then used to guide a simple demosaicking method based on graph regularization.

Image Interpolation with Contour Stencils

Pascal Getreuer

Computer Vision and Pattern Recognition · Image Processing On Line · 31 Jul 2011 · cited 20× · DOI ↗ · more like this ·

Image interpolation is the problem of increasing the resolution of an image. Linear methods must compromise between artifacts like jagged edges, blurring, and overshoot (halo) artifacts. More recent works consider nonlinear methods to improve interpolation of edges and textures. In this paper we apply contour stencils for estimating the image contours based on total variation along curves and then use this estimation to construct a fast edge-adaptive interpolation.

Zhang-Wu Directional LMMSE Image Demosaicking

Pascal Getreuer

Computer Vision and Pattern Recognition · Image Processing On Line · September 2011 · cited 19× · DOI ↗ · more like this ·

Most digital cameras capture samples through a color filter array. At every pixel location, the camera observes one of either the red, green, or blue component. Demosaicking (or demosaicing) is the problem of using this incomplete information to estimate all three color components at every pixel. Zhang and Wu proposed an effective solution to this problem in “Color Demosaicking via Directional Linear Minimum Mean-Square-Error Estimation”.

Image Interpolation with Geometric Contour Stencils

Pascal Getreuer

Computer Vision and Pattern Recognition · Image Processing On Line · September 2011 · cited 18× · DOI ↗ · more like this ·

We consider the image interpolation problem where given an image v<sub>m,n</sub> with uniformly-sampled pixels vm,n and point spread function h, the goal is to find function u(x,y) satisfying v<sub>m,n</sub> = (h*u)(m,n) for all m,n in Z. This article improves upon the IPOL article Image Interpolation with Contour Stencils. In the previous work, contour stencils are used to estimate the image contours locally as short line segments. This article begins with a continuous formulation of total variation integrated over a collection of curves and defines contour stencils as a consistent discretization. This discretization is more reliable than the previous approach and can effectively distinguish contours that are locally shaped like lines, curves, corners, and circles. These improved contour stencils sense more of the geometry in the image. Interpolation is performed using an extension of…

Color demosaicing with contour stencils

Pascal Getreuer

Computer Vision and Pattern Recognition · July 2011 · cited 17× · DOI ↗ · more like this ·

Demosaicing is the problem where, given a color image that has been subsampled by a color filter array, to interpolate the complete color information at each pixel. Many demosaicing methods adapt the interpolation according to estimated edge orientations. However, while the accuracy of these orientations is important for success, they are difficult to estimate from the mosaiced image. This paper extends contour stencils, a method for estimating image contours based on total variation along curves, to estimating contour orientations directly on mosaiced images. A method for demosaicing using these contour orientation estimates is proposed and compared with existing methods. The proposed demosaicing is performed as an energy minimization, using a graph regularization adapted according to the orientation estimates.

ENO Multiresolution Schemes with General Discretizations

Pascal Getreuer, François G. Meyer

Computer Vision and Pattern Recognition · SIAM Journal on Numerical Analysis · 2008 · cited 13× · DOI ↗ · more like this ·

Harten's framework is a nonlinear generalization of the wavelet framework. Previously, the choice of discretization (scaling function) in Harten multiresolution schemes has been limited to point-value, cell-average, and hat-based discretization. This paper shows how to construct multiresolution schemes consistent with Harten's framework for a variety of discretizations. The construction here begins with the discrete operators and deduces the corresponding continuous operators, reversing the order of the usual approach. This construction yields as a special case essentially nonoscillatory (ENO) multiresolution schemes for any order of spline discretization and also has the flexibility to define multiresolution schemes with nonspline discretizations. An error-control strategy is also developed.

Image zooming with contour stencils

Pascal Getreuer

Computer Vision and Pattern Recognition · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 12 Jan 2009 · cited 12× · DOI ↗ · more like this ·

We introduce "contour stencils" as a simple method for detecting the local orientation of image contours and apply this detection to image zooming. Our approach is motivated by the total variation along curves: small total variation along a candidate curve suggests that this curve is a good approximation to the contours. Furthermore, a relationship is shown between interpolation error and total variation. The contour stencil detection is used to develop two image zooming methods. The first one, "contour stencil interpolation," is simple and computationally efficient, yet competitive in a comparison against existing methods. The second method approaches zooming as an inverse problem, using a graph regularization where the graph is determined by contour stencil detection. Both methods extend naturally to vector-valued data and are demonstrated for grayscale and color images.

Malvar-He-Cutler Linear Image Demosaicking

Pascal Getreuer

Computer Vision and Pattern Recognition · Image Processing On Line · 14 Aug 2011 · cited 10× · DOI ↗ · more like this ·

Image demosaicking (or demosaicing) is the interpolation problem of estimating complete color information for an image that has been captured through a color filter array (CFA), particularly on the Bayer pattern. In this paper we review a simple linear method using 5 x 5 filters, proposed by Malvar, He, and Cutler in 2004, that shows surprisingly good results.

Local Kernels that Approximate Bayesian Regularization and Proximal Operators

Frank Ong, Peyman Milanfar, Pascal Getreuer

cs.CV · 9 Mar 2018 · cited 8× · arXiv ↗ · more like this ·

In this work, we broadly connect kernel-based filtering (e.g. approaches such as the bilateral filters and nonlocal means, but also many more) with general variational formulations of Bayesian regularized least squares, and the related concept of proximal operators. The latter set of variational/Bayesian/proximal formulations often result in optimization problems that do not have closed-form solutions, and therefore typically require global iterative solutions. Our main contribution here is to establish how one can approximate the solution of the resulting global optimization problems with use of locally adaptive filters with specific kernels. Our results are valid for small regularization strength but the approach is powerful enough to be useful for a wide range of applications because we expose how to derive a "kernelized" solution to these problems that approximates the global…

Image stylisation: from predefined to personalised

Ignacio Garcia‐Dorado, Pascal Getreuer, Bartlomiej Wronski +1 more

Computer Vision and Pattern Recognition · IET Computer Vision · 3 Apr 2020 · cited 6× · DOI ↗ · more like this ·

The authors present a framework for interactive design of new image stylisations using a wide range of predefined filter blocks. Both novel and off‐the‐shelf image filtering and rendering techniques are extended and combined to allow the user to unleash their creativity to intuitively invent, modify, and tune new styles from a given set of filters. In parallel to this manual design, they propose a novel procedural approach that automatically assembles sequences of filters, leading to unique and novel styles. An important aim of the authors’ framework is to allow for interactive exploration and design, as well as to enable videos and camera streams to be stylised on the fly. In order to achieve this real‐time performance, they use the Best Linear Adaptive Enhancement (BLADE) framework – an interpretable shallow machine learning method that simulates complex filter blocks in real time.…

Gunturk-Altunbasak-Mersereau Alternating Projections Image Demosaicking

Pascal Getreuer

Computer Vision and Pattern Recognition · Image Processing On Line · September 2011 · cited 3× · DOI ↗ · more like this ·

The problem of image demosaicking (or demosaicing) is where an image has been captured through a color filter array (CFA), and the goal is to estimate complete color information at every pixel. This IPOL article describes the image demosaicking method proposed by Gunturk, Altunbasak, and Mersereau in "Color Plane Interpolation Using Alternating Projections." Given an initial demosaicking, the method improves the result by alternatingly applying two different projections. One projection copies the green channel's wavelet detail coefficients to the red and blue channels while the other projection constrains the solution to agree with the observed data.

Image Stylization: From Predefined to Personalized

Ignacio Garcia-Dorado, Pascal Getreuer, Bartlomiej Wronski +1 more

cs.GR · 22 Feb 2020 · cited 2× · arXiv ↗ · more like this ·

We present a framework for interactive design of new image stylizations using a wide range of predefined filter blocks. Both novel and off-the-shelf image filtering and rendering techniques are extended and combined to allow the user to unleash their creativity to intuitively invent, modify, and tune new styles from a given set of filters. In parallel to this manual design, we propose a novel procedural approach that automatically assembles sequences of filters, leading to unique and novel styles. An important aim of our framework is to allow for interactive exploration and design, as well as to enable videos and camera streams to be stylized on the fly. In order to achieve this real-time performance, we use the \textit{Best Linear Adaptive Enhancement} (BLADE) framework -- an interpretable shallow machine learning method that simulates complex filter blocks in real time. Our…

Continuous Optimization for p-adic Models

Julian Salazar, Dimitri Kanevsky, Matt Harvey +2 more

cs.LG · 21 Sep 2026 · cited 0× · arXiv ↗ · more like this ·

We present the first method for native, continuous gradient descent for machine learning models with $p$-adic parameters. Existing native optimizers are discrete, mostly combinatorial searches, as the $p$-adic numbers $\mathbb{Q}_p$ are totally disconnected, with standard losses that are flat away from their minima. To enable continuous optimization, we propose working with $\mathbb{Q}_p$ via its Berkovich affine line: a canonical, path-connected expansion of $\mathbb{Q}_p$ that preserves its isometries and uniquely extends its analytic maps. This hull is a metric tree with interpretable points and local derivatives, which we show enables effective optimizers and backpropagation. We formulate gradient descent and show that its approximations efficiently learn linear models with coefficients in $\mathbb{Q}_p$ to do modular arithmetic, an XOR-like task not expressible by linear models in…

Fast, Trainable, Multiscale Denoising

Sungjoon Choi, John Isidoro, Pascal Getreuer +1 more

cs.CV · 16 Feb 2018 · cited 0× · arXiv ↗ · more like this ·

Denoising is a fundamental imaging problem. Versatile but fast filtering has been demanded for mobile camera systems. We present an approach to multiscale filtering which allows real-time applications on low-powered devices. The key idea is to learn a set of kernels that upscales, filters, and blends patches of different scales guided by local structure analysis. This approach is trainable so that learned filters are capable of treating diverse noise patterns and artifacts. Experimental results show that the presented approach produces comparable results to state-of-the-art algorithms while processing time is orders of magnitude faster.

“Color Demosaicking via Directional Linear Minimum Mean-Square-Error Estimation ” [4]. Exact Observed Image Zhang-Wu Demosaicked Image Demosaicking example. The images are enlarged to show individual pixels more clearly. References

Pascal Getreuer, Pascal Getreuer, Cmla Ens Cachan +5 more

Computer Vision and Pattern Recognition · 8 Oct 2016 · cited 0× · OpenAlex ↗ · more like this ·