J.-l. Starck - Academia.edu (original) (raw)

Papers by J.-l. Starck

Spectral matching ICA (SMICA) is a source separation method based on covariance matching in Fouri... more Spectral matching ICA (SMICA) is a source separation method based on covariance matching in Fourier space that was designed to address in a flexible way some of the general problems raised by Cosmic Microwave Background data analysis. However, a common issue in astronomical data analysis is that the observations are unevenly sampled or incomplete maps with missing patches or intentionally masked parts. In addition, many astrophysical emissions are not well modeled as stationary processes over the sky. These effects impair data processing techniques in the spherical harmonics representation. This paper describes a new wavelet transform for spherical maps and proposes an extension of SMICA in this space-scale representation.

anisotropy) is vitally important. This suite of new techniques allows the accurate extraction of ... more anisotropy) is vitally important. This suite of new techniques allows the accurate extraction of phase space densities with fidelity impossible to achieve with naive or trivial interpolation or smoothing techniques. We will demonstrate the relative strengths and detection capabilities of these techniques. The examples we treat are the Beam-Plasma Instability (BPI) and ponderomotively driven KEEN waves. These allow intricate phase space structures to coexist with chaos and turbulence. The former has a simple single mode linear limit which is unstable, but the latter gives rise to pure multimode nonlinear phenomena.

Journal of Cosmology and Astroparticle Physics, 2015

Astronomy and Astrophysics Library, 2006

Statistical Challenges in Modern Astronomy II, 1997

We describe in this paper the Multiscale Maximum Entropy Method which is based on the concept of ... more We describe in this paper the Multiscale Maximum Entropy Method which is based on the concept of multiscale entropy derived from the wavelet decomposition of a signal into different frequencies bands. It leads to a method which is flux conservative, and the use of a multiresolution support solves the problem of MEM to chose the a parameter, i.e. relative weight between the goodness-of-fit and the entropy.

We present the data for the LRG sample used, including the value of Wmax. Wmax is the value of th... more We present the data for the LRG sample used, including the value of Wmax. Wmax is the value of the wavelet coefficient of the transformation for the values of the parameters R (radius) and s (width) that give the maximum of the global statistic B(R,s). In this way, the values Wmax are a measure of how strong is the signal coming from a BAO shell around a given LRG. (1 data file).

Proceedings of SPIE - The International Society for Optical Engineering, 2002

One of the goals in cosmology is to understand the formation and evolution of the structures resu... more One of the goals in cosmology is to understand the formation and evolution of the structures resulting from the growth of initial density perturbations. Recent Cosmic Microwave Background (CMB)observations indicate that these pertubations essentially came out of Gaussian distributed quantum fluctuations in the inflationary scenario. However, topological defects (e.g. cosmic strings) could contribute to the signal. One of their important footprints would be the predicted non-Gaussian distribution of the temperature anisotropies. In addition, other sources of non-Gaussian signatures do contribute to the signal, in particular the Sunyaev-Zel'dovich effect of galaxy clusters. In this general context and motivated by the future CMB experiments, the question we address is to search for, and discriminate between, different non-Gaussian signatures. To do so, we analyze simulated maps of the CMB temperature anisotropies using both wavelet and curvelet transforms. Curvelets take the form of basis elements which exhibit very high directional sensitivity and are highly anisotropic, which is not the case for wavelets. The sensitivity of both methods is evaluated using simulated data sets.

Massive Computing, 2002

With the requirements of scientific and medical image database support in mind, we describe a ran... more With the requirements of scientific and medical image database support in mind, we describe a range of useful technologies for storage, transmission and display. These new technologies are all based on discrete wavelet or related multiscale transforms. Other important issues include noise modeling, and the innovative use of entropy for information characterization.

Signal Processing, 2001

ABSTRACT

Journal of the Optical Society of America A, 1992

The basic concepts of optical interferometric imaging through the atmosphere at a low light level... more The basic concepts of optical interferometric imaging through the atmosphere at a low light level are applied to the case of the Calern High Angular Resolution Optical Network (CHARON) stellar interferometric array. The numerical simulation that was implemented to create the interferometric data is presented. The processing algorithms used to process the raw data, to extract the object parameters, and to restore the initial map are pointed out. The multiresolution approach provides an objective way of analyzing the reconstruction procedure. Reconstructed maps under different conditions of brightness and turbulence are shown and discussed. The advantages and the drawbacks of the different steps of the computer simulation are analyzed.

The Multiscale Vision Model is a recent object detection method, based on the wavelet transform. ... more The Multiscale Vision Model is a recent object detection method, based on the wavelet transform. It allows us to extract all objects contained in an image, whatever their size or their shape. On each extracted object, information concerning its flux or its shape can easely be determined. We show that such an approach can be combined with a deconvolution, leading to the reconstruction of deconvolved objects. We discuss the advantages of this method, such the way to perform a deconvolution with a space variant Point Spread Function. We present a range of examples and applications to illustrate the effectiveness of this approach.

Statistics and Computing, 2000

We describe a range of powerful multiscale analysis methods. We also focus on the pivotal issue o... more We describe a range of powerful multiscale analysis methods. We also focus on the pivotal issue of measurement noise in the physical sciences. From multiscale analysis and noise modeling, we develop a comprehensive methodology for data analysis of 2D images, 1D signals (or spectra), and point pattern data. Noise modeling is based on the following: (i) multiscale transforms, including wavelet

Computing in Science & Engineering, 2002

In this work we investigate a new approach to remove the bright compact source contribution to th... more In this work we investigate a new approach to remove the bright compact source contribution to the Cosmic Microwave Background (CMB). Bright compact source emissions contaminate the full-sky CMB data over a significant fraction of the sky. Besides, at small scales (typically multipoles greater than 2000), they represent the dominant foreground contribution to the CMB. These emissions should therefore be removed from CMB data if a full-sky estimate of the components is sought after. However their spectral variability makes difficult to blindly separate them from other emissions, even using the recent stateof-the-art localized source separation techniques. To date, after detection, the brightest sources are either masked and inpainted prior to CMB analysis, or their flux is estimated by minimizing a local χ 2 with a background assumed constant. In this work, we rather propose to estimate the flux of the brightest compact sources in the CMB data using a morphological separation approach, including a less crude model for the background. We propose to separate compact sources with know support and shape from a background assumed sparse in the spherical harmonic domain. Results supporting this approach are presented on full-sky simulations of compact source contributions to CMB maps based on the ERCSC catalogue [1].

The spectral energy distribution of HH 100 IRS is modeled as a function of dust parameters such a... more The spectral energy distribution of HH 100 IRS is modeled as a function of dust parameters such as the grain size, the ice volume fraction, and the fluffiness of the particles. The radiative transfer calculations include a detailed treatment of the spectroscopic signature of the ice bands. The infrared spectrum of HH 100 IRS is successfully reproduced if an upper