Position initialization on brightest central object?

[aagnello]

Some cutouts may be offset... so that choosing the central pixel of image 1 may not be a good starting point for all other images. First: check how centroid initislisation is currently done, and then maybe switch to a flux-weighted centroiding. Don't forget, tractor objects have positions in world coordinates...

[aagnello]

PSEUDOcode for recentering. Not pythonic syntax all the way through,
e.g. in the SUM commands, which must be replaced with their python
counterparts. Should be refactored in separate module/class/method...
to be used in the three initialisations.
NB: At the moment the center coordinates are the same for
all cards. While ok (and more convenient than starting blindly at
NX/2, NY/2), different cards may be offcentered in different manners.
Task for future(!) refurbishing: allow for different centroids in
different cards, but with same gal-img displacements.
Here it is:

sci = images[0].data
magicrec = 20
magicwx = int(NX/2)
magicwy = int(NY/2)
#first initialise xc,yc to either x,y or xs,ys! then:
irec = 1
while (irec<magicrec):
tfl =
SUM[sci[ix,iy]_BOOLE[abs(ix-xc)<magicwx]_BOOLE[abs(iy-yc)<magicy],{ix,1,NX},{iy,1,NY}]
xrec =
SUM[ix_sci[ix,iy]_BOOLE[abs(ix-xc)<magicwx]_BOOLE[abs(iy-yc)<magicy],{ix,1,NX},{iy,1,NY}]/tfl
yrec =
SUM[iy_sci[ix,iy]_BOOLE[abs(ix-xc)<magicwx]_BOOLE[abs(iy-yc)<magicy],{ix,1,NX},{iy,1,NY}]/tfl
xc = int(xrec)
yc = int(yrec)
irec++

#possibly apply the small displacemnts to initialise pt.sources, then:
centerpos = wcs.pixelToPosition(xc,yc)

[drphilmarshall]

OK, flux-weighted centroids are computed in lenstractor.Deal (dm.py) and returned; the mean ra and dec are then used as the initial Nebula galaxy position.