Concept
Anisotropic diffusion
Publications associées (57)
Segmentation of Natural Images Using Scale-Space Representations: A Linear and a Non-Linear Approach
Pierre Vandergheynst, Oscar Divorra Escoda
In general purpose computer vision systems, unsupervised image analysis is mandatory in order to achieve an automatic operation. In this paper a different approach to image segmentation for natural scenes is presented. Scale-Space representation is used to ...
IEEE2002Properties of Smoothing with Time Gating
Estimates of transfer functions from measurements contain undesired information -- noise -- whose effect can partly be taken away by applying different smoothing techniques. One such smoothing technique suggests to remove the noise by gating the impulse re ...
2000Segmentation and symmetry measure for image analysis
In this thesis, four topics related to image analysis are investigated: filtering, color segmentation, contour detection, and symmetry quantification for shape, color, and texture information. The three first topics are essential to detect efficiently the ...
EPFL1999Analysis of Rainfall Data by Robust Spatial Statistics using S+SpatialStats
This paper discusses the use of robust geostatistical methods on a data set of rainfall measurements in Switzerland. The variables are detrended via non-parametric estimation penalized with a smoothing parameter. The optimal trend is computed with a smooth ...
1998Aggregation of Fractal and Dendritic Ag Clusters on a Pt(111) Surface
We have used variable-temperature scanning tunneling microscopy to study the aggregation of two-dimensional Ag clusters on Pt(111). A transition from randomly ramified to dendritic fractal growth is observed in the diffusion-limited regime. Atomic-scale ob ...
1995Mechanism of the Transition from Fractal to Dendritic Growth of Surface Aggregates
THE similarity of many patterns formed in non-equilibrium growth processes in physics, chemistry and biology is conspicuous, and many attempts have been made to discover common mechanisms underlying their formation(1). A central question is what causes som ...
1994Simulation of 1D ARPES Data and Measures to recover the underlying Signal
Fourier deconvolution is a good tool to remove the blurring of images with a very high SNR. However it suffers under severe noise degradation and becomes impracticable at a SNR of 20dB. If the convolution width is very high this happens even earlier. The a ...
0