#AUTHOR: Bekeraite, Walcher
#SOURCE: SDSS DR7 images and own analysis
#DATE: 2014-09-09
#VERSION: 1.0
#COLAPRV: CJW
#PUBAPRV: none

This table provides the results of a reanalysis of the SDSS DR7 images 
and the derivation of growth curve magnitudes (GCmag). 

SExtractor was used to detect stars and other sources in the image to 
create a pixel mask for all the sources except the CALIFA galaxy. The 
mask was visually checked and manually corrected, if necessary.

Simple interpolation would not by definition have been applicable, 
so we chose a procedure known as inpainting in the computer vision 
community, i.e. masked pixels were iteratively replaced with a Gaussian 
inverse-distance weighted average of the neighbouring real pixels, 
starting with the pixels with the largest number of nearest neighbours. 
In order to apply the masks (originally created for the r band images) 
to the other 4 SDSS bands, we measured the shift between the different 
images and their r band counterparts using their WCS (FITS World 
Coordinate System) ra and dec coordinates, then shifted the masks and 
cropped the images accordingly.
                                   
The position angle and ellipticity of the galaxies were measured based 
on the light moments of the masked galaxy image (compare the SExtractor 
manual for definitions of the light moments). 

The growth curve analysis was performed in both elliptical and circular 
(r-band only) incrementally increased annuli.


If the flux profile was fitted in sufficiently wide rings using simple
linear regression, the best fit line should become horizontal at some 
radius, which might then be considered the edge of the galaxy. This would 
assume that flux falls off asymptotically until it is indistinguishable 
from the sky fluctuations. In practice this is not the case, given that 
incomplete masks, light from other objects and sky gradients make the 
best fit slope switch from negative to slightly positive at some point. 
We therefore fit 150 px sections of the flux profile using simple linear 
regression, making the neighbouring fit sections overlap by 100 
pixels. When the flux profile slope becomes non-negative, we take the 
mean of the current ring as the sky value, and the ellipse with major 
axis value at the middle of the ring as the galaxy's edge.

After measuring the galaxy's extent and sky value, the flux within the 
galaxy was summed up, sky was subtracted and u, g, r, i, z magnitudes 
were calculated.

The counts were converted to magnitudes, as described here:
http://www.sdss.org/dr7/algorithms/fluxcal.html#counts2mag
87 galaxies had no zeropoint/extinction/airmass given in corresponding 
file headers, in such a case the mean values of these parameters in a 
given band was used.


The magnitudes were converted to AB magnitude system by applying 
zeropoint correction as described in 
http://www.sdss3.org/dr8/algorithms/fluxcal.php#SDSStoAB. The correction 
is as follows: u_{AB} = u_{SDSS} - 0.04 mag
The unconfirmed z band correction, z_{AB) = z_{SDSS} + 0.02 mag, was 
not applied. 

The half light radii (semi-major axes for the elliptical case) were 
obtained by taking the semi-major axis value of an ellipse such that 
half of the total flux lies inside or on it, and multiplying it by the 
SDSS pixel scale (0.396 "/pix).

A set of quality flags were also introduced after a visual inspection, 
denoting uncertain position angle values, incorrect center coordinates 
(a significant offset), unreliable measurements. The last column in the 
table is the sum of quality bit flags. The meaning of the individual 
flags are:

-- 1: Visually wrong PA
-- 2: Visually wrong center (large offset)
-- 4: Unreliable photometry: bad masking, distorted shapes due to 
inpainting, too close to a frame border, weird shape of the growth curve.