4.1.1 Finding and Classifying Objects

The first problem to address is what to do with all the pixels in the image that are contaminated by signal from other `objects', such as other galaxies, stars, cosmic-ray-events, and remanence or overflow stripes that could not be removed. In our scheme, all such pixels are simply not used in the fit, with the exception of severely overlapping galaxies, cf. below.

The problem is then how to determine which pixels should be flagged (i.e. omitted from the fit). We do this as follows. First we attempt to automatically find all objects in the image using daofind called with suitable parameters. Based on the output from daofind (peak value, roundness and sharpness) and on the seeing value and the CCD saturation level we then classify the found objects as either `stars' or cosmic-ray-events. The star category includes all objects that extend over several pixels and are `unsharp', i.e. stars and galaxies. The star objects are assigned a flagging radius based on their estimated extend. The cosmic-ray-event objects are assigned a flagging radius of 1 pixel. In the fits, pixels within a circle centered on each object and with the corresponding radius are omitted.

The total number of objects found is in the range 400-1000 per image. This number depends on the exposure time (since the number of cosmic-ray-event grows linearly with exposure time), the fields looked at (how many stars and galaxies there actually are in the image), and the seeing (since whether objects with small angular separation can be separated depends on the seeing). The point here is just that the number is so high that an automated procedure is clearly needed. However, the algorithm that assigns radii to the `stars' sometimes fails, especially for `stars' located close to the center of the galaxy we want to fit where the background is not flat. The user can then intervene and manually change the radii.