8 Examining image flatness

One of the major problems encountered when using astronomical images obtained from CCDs or plate scans is that of poor flatfielding. Very often we are forced to deal with images that have been pre-processed by someone else and yet still retain variations in the background value. These are most commonly caused either by hardware faults such as vignetting, poor flatfielding software or by simple human error.

Whilst such variations may be small compared to the brighter objects on an image, they can make analysis of faint and/or diffuse objects very difficult and, in addition, they will spoil any low level display of the image by appearing as brighter or darker regions rather than a uniform background continuum.

The more serious quantitative influence of these variations is in the consequent inaccuracy of any global background value estimate fed into a profiling application. If a galaxy being profiled is in a part of the image for which the global background value is significantly inaccurate the profile generated will also be inaccurate. Any subsequent scale length analysis will also be wrong, particularly at lower isophotes.

The ESP application SKEW allows you to generate an image which highlights where the flatfielding has been poorly accomplished. It operates by considering how the skewness of the pixel brightness distribution varies over the image. It does this by, for each pixel of the image in turn, considering the values of all pixels within a given radius and using those to calculate a skewness value. You must define the size of the image region sampled and hence the scale size over which fluctuations in the background may be detected. Any areas of the image where the pixel brightness distribution is not similiar to the expected Normal distribution are shown up strongly. The values for the expected distribution are taken from the modal value and background value standard deviation of the image. These may be derived easily using HISTPEAK.

For images containing very bright galaxies it may be necessary to reduce the contribution of real objects in the image by use of the TOPPED application. This then allows the variations in the background to be more easily seen.

It should be noted that SKEW is not a quick application since a large image and sampling area can easily combine to cause the application to make billions of calculations per run. However, it is a good idea to run the application at least once for every new data set unless you are entirely confident of the flatfielding techniques that were employed during reduction. Execution time may be reduced by reducing the size of an image using KAPPA’s COMPADD prior to using SKEW.

The Starlink package CCDPACK is particularly useful for flatfielding data even when no flat frames are available.

The SKEW application can be used with the following syntax:

  % skew in=jet out=skewim modet=true psize=0.5 width=10. mult=1000.
               back=949 useall=true

This example leads to a skew type image named skewim being generated. The image’s global modal pixel value (949) is employed and all the image pixels are allowed to contribute to the skewness calculations (best used only if no bright objects are present). The sampling area used for the calculation will be WIDTH arc seconds (=WIDTH/PSIZE pixels) square. The final skewness values calculated are multiplied by the factor 1000 (making them more acceptable to HISTPEAK). As usual, if the image contains a SKY co-ordinate frame then PSIZE need not be speficied. Other examples may be found in Appendix 0.