At the last ImageDetWG meeting there was some discussion at the end regarding what morphological parameters to include in the database, but the full discussion was postponed for next time. The discussion centered around concentration index and asymmetry parameter, as I recall. Are these two the only ones under consideration?
Here is a short list of morphological parameters that have been used in galaxy (and other) image processing and classification: 1.. Concentration Index C(alpha): If image has radius R then C(alpha) = Sum I(alpha x R) / Sum I(r). Typically a value of alpha = 0.3 has been adopted (Abraham et al, 1994; Naim et al., 1995) 2.. Concentration Index C = 5 log r(80%)/r(20%): (Bershady et al., 2000; Conselice 2003) where r(80%) is the image radius containing 80% of the light from the galaxy and similarly for r(20%). 3.. Asymmetry Parameter A(theta) = abs(I - I(theta))/2 abs(I): (Conselice et al., 2000) Generally used with theta = 180 degrees or sometimes 90 degrees. 4.. Ellipticity: Ratio of semi-major to semi-minor axes. Depends on aspect angle. 5.. max(rI): Maximum of the plot of rI(r) vs. r. 6.. r(25)/r(75): ratio of the radii at which 25% and 75% of light is enclosed in a plot of I(r) vs. r. 7.. Blobbiness, Isophotal displacement, Isophotal filling factor, structure skeletons: These are parameters described by Naim et al., 1997. They basically tried to come up with some parameters that measured the texture and smoothness of galaxies. 8.. There are measures of Entropy and energy of an image that are used in Earth science imagery to measure the texture of an image. These are based on the co-occurrence matrix, Cij, which is a measure of probability that a pixel with value i has a neighboring pixel of value j. Entropy = -Sum(i,j) Cij log Cij. Several other parameters that are derived from Cij are also used as measures of texture. I am sure there are others. Any input on what are the benefits of each of these? How many should be included in the database? Dave Bazell _______________________________________________ LSST-data mailing list [email protected] http://www.lsstmail.org/mailman/listinfo/lsst-data
