“Probabilistic Analysis for Space & Earth Observations”
Multisource & Multispectral Image Processing via Bayesian Inference 		MIV team LSIIT

CNRS – UDS
Galaxy classification using both spatial (morphological) and spectral aspects
Title : GalaxyClassification
Keywords : galaxy classification, multivariate cube, decomposition into structures, Bayesian inference, mathematical morphology
Funding : MENRT
Start date : 2007-10-01
End date : 2010-10-01
  • People
  • Summary
  • Description
  • Bibliography
  • Links

    • People

    Participants

    Collaborators

    Summary

    At present, the galaxies are classified using their morphological visual appearance (e.g. Hubble sequence). However, studying the galaxies at several wavelength is very useful to understand their dynamics, their evolution, and, consequently, the universe history. Therefore, the goal of the project is to propose a classification method using both the spatial and spectral informations. We propose to use two complementary approaches in image processing: Bayesian inference and mathematical morphology. The first one considers a marked point process model set in a Bayesian framework, from which a posterior distribution is optimized using MCMC method to estimate the major geometrical parameters of the galaxy (center, size, presence of a bar, etc.). Then mathematical morphology provides informations about spiral arms, distribution of stars, ...

    Description

    Key objectives

    Quantitative morphology is a popular approach for monoband galaxy analysis. It produces accurate and robust automatic classification in the Hubble scheme. This project aims at developing methods for the automatic extraction of quantitative spectro-morphological features from multispectral galaxy observations in local universe.
    The chosen approach is based on parametric galaxy modeling. In this model, a galaxy is described as the composition of several possible structures: the bulge, the stellar disc, the stellar bar, the spiral arms, the dust lane, the stellar ring, ... Each structure is physically motivated and is modeled with parameters of high-level semantic (geometry, brightness, color) leading to a straightforward interpretation. The extraction of this quantitative spectro-morphological description from a large set of galaxies will enabled us to build a new comprehensive and objective multispectral classification of galaxies.

    VisuTool

    VisuTool is a Java application to demonstrate the technique of parametric galaxy modeling. It simulates the appearance of a galaxy having the desired parameters (shape, brightness) and it is able to simulate the effect of the optics, the atmosphere (noise, PSF). The demonstration version uses a model composed of a bulge, a stellar disc, a stellar bar and spiral arms with a Gaussian PSF and Gaussian noise. All parameters from all structures can be (almost) freely modified and the effect of the modifications are visualized in real time in the simulation.
    A demonstration version of VisuTool can be launched directly from your web browser using Java Web Start : Click here to launch the application! An application run under Java Web Start works in a sand box and cannot access to your system, thus you won't be able to save/load files... Java Web Start is included in Java Sun JRE, you can download it at : Sun Web Site

    Quick Video Tutorial

    Here is a quick Flash tutorial to teach you basic uses of VisuTool :



    This second video shows some advanced tuning options:



    Quick Text Tutorial

    The application is composed of 2 frames. The main one is itself divided into 2 parts.

    Oups there should be an image here!


    The left part represents the model tree, and the right one shows the galaxy and its different structures. You can expand the different sections of the tree to access the variables of each structure. Then, you can modify the value of the parameters and directly see how it affects the galaxy in the right part. The right part is a multiple image visualizer. You can select the image you want to see in the right list. There are several thing you can do like zoom in and out with the mouse wheel or change the histogram correction option by holding Alt key and playing with mouse wheel...

    The second frame shows a profile cut of the galaxy and its components through the red line drawn in the main frame.

    Oups there should be an image here!


    Each line represents a structure and the vertical lines represent the position of the effective radius of the structure of the same color. You can move the red line with the mouse in the main frame. Just click on an extreme point of line and drag it.

    Bibliography

    Links


    Webmaster (M. Petremand)
    © PASEO Group 2005 | WebSite Info & Credits
    Last update: Jan 04, 2010
    PASEO
    Research Group
    Home
    Research projects
    Publications
    Software
    Teaching materials
    News & Events
    Seminars

    People
    Collaborations
    Job Openings
    Links & Resources
    Wiki

    Local Admin