Re: [TASS] Flat fielding

[mgutzwiller@LANVISION.COM]

The current version of the Flat program is not affected by changing sky
brightness because it uses a slightly different algorithm.  This is how it
works:

1.      An ordinary image, preferably one with a high background level, is
dark subtracted.
2.      The median value of each row is calculated.
3.      Each pixel in the row is replaced by the ratio of its value to the
row's median to create a scaled image.
4.      The median value of each column in the scaled image is calculated to
create a flat vector.
5.      The scaled flat vector is written to a FITS image file where the
values are scaled to fit the 2 byte integer range.

This algorithm avoids the problem with changes in the sky brightness and
also reduces its sensitivity to bright stars.


Mike G.

                -----Original Message-----
                From:   Andrew Bennett
[mailto:andrew.bennett@NS.SYMPATICO.CA]
                Sent:   Wednesday, September 08, 1999 8:22 PM
                To:     TASS@LISTSERV.WWA.COM
                Subject:        Flat fielding

                In TN0019 (960903) Michael Richmond wrote:
                > I made a flatfield vector for each segment by
                >calculating the median value of pixels in each
                >column. I then divided each image by this 1-D
                >vector. After looking at some of the reduced
                >images, it appears that I'll need to find a
                >better way of calculating the flatfield vector;
                >this method can be screwed up by bright stars and
                >changing sky brightness.

                To which Herb Johnson in TN0057 (Aug 19 1999) added
                >[The image should be a twilight image of a
                >high and relatively uniform illumination]

                Ah! It should be. But is it? Are all the
                flatfield images up to standard? Or could this
                be a major contributor to the 0.05 magnitude
                noise floor?

                In other contexts, the importance is stressed
                of averaging as many flatfield determinations
                as possible. As far as I can see, our software
                does not allow for averaging. Should it?

                Andrew Bennett, Avondale Vineyard, Nova Scotia, Canada.