7.10.1: Appendix, Image Sections for Generalized Binning

This specifies the general case of how to document which pixels are which in binned mosaicked images. IRAF-style image sections of binned images

This appendix attempts to cover the issues involved in specifying exactly which region of silicon on the CCD was stored into a FITS image array. The proposed FITS WCS conventions are used to denote the coordinate system on the silicon. Pixels are numbered starting with 1, and the center of a pixel has integral location.

For the sake of this example, let us presume a detector which is 1024x1024 pixels. Unbinned readout

For an unbinned readout of the entire CCD in NOAO notation this would have
DATASEC = '[1:1024,1:1024]'
CCDSEC  = '[1:1024,1:1024]'
CCDSUM  = '1 1'
ORIGSEC = '[1:1024,1:1024]' Simple binning by 2

Now say that the entire detector was read out binned by 2 in each direction. I do not know for certain how NOAO data acquisition systems define ORIGSEC in the case of binned readouts, but it seems to make most sense for it to refer to the largest possible DATASEC of the binned detector. I will use that convention here.
DATASEC = '[1:512,1:512]'
CCDSEC  = '[1:512,1:512]'
CCDSUM  = '2 2'
ORIGSEC = '[1:512,1:512]'
But if only the quarter of the CCD farthest from the amplifier were stored in the FITS image.
DATASEC = '[1:256,1:256]'
CCDSEC  = '[257:512,257:512]'
CCDSUM  = '2 2'
ORIGSEC = '[1:512,1:512]' Atypical binning by 2

When a detector is read out in a binned fashion it is possible that the readout window may exclude a number of physical pixels which is not evenly divided by the binning factor. (This could be required by certain constraints in the CCD or controller design.) Say that in the case of the detector above the first and last row and column of the detector were not within the readout region. Within the IRAF nomenclature the a way to represent this would be.
DATASEC = '[1:511,1:511]'
CCDSEC  = '[1.5:511.5,1.5:511.5]'
CCDSUM  = '2 2'
ORIGSEC = '[1:512,1:512]'
I am not certain that any IRAF tasks know how to handle non-integral values of the array limits. I see no better way of communicating the readout window of the CCD. Generalized rules for binning

In the general case of binning the following equations describe the spatial extent of a particular pixel.

Along any axis let the binning factor be b. Let the binned pixel number, as would be inserted in a CCDSEC card, be n. Let the total number of pixels along the axis be M. The binned pixel number n is permitted to take on any value which is an integral multiple of 1/b. Values of n from 1 to M (inclusive) indicate binned pixels which consist entirely of actual imaging pixels on the CCD. Values of n outside those bounds indicate binned pixels which contain some amount of prescan or overscan data. Values of n between 0 and 1 indicate that the bin begins before the edge of the imaging area of the CCD. Values of n between M and M+1 indicate that the bin ends beyond the edge of the imaging area of the CCD. When n is 0 or less or when n is M+1 or greater the binned pixel is entirely outside the imaging array.

The coordinates given below are in terms of the unbinned CCD pixels.

The minimal extent of pixel n is located at coordinate 0.5 + (n - 1) * b.

The middle of pixel n is located at coordinate 0.5 + (n - 0.5) * b.

The maximal extent of pixel n is located at coordinate 0.5 + n * b.

Steve Allen <sla@ucolick.org>
$Date: 1996/03/18 23:44:25 $