Source Detection

Command:

sou_main

Scope:

Detect discrete sources and calculate both significance and count rate

Algorithm:

Wavelet detection - using the FFT and the Mexican cap filter on

various scales. Exposure map is explicitly used. The ratio of the wavelet

image and its uncertain image is used as the S/N ratio map. Local S/N maxima

are searched and located as candidates of sources. This routine is

fast and dirty. The background is locally determined, thus subjecting to large

counting uncertainties of the background regions.

Map detection - a sliding box method with the background determined

on larger scales. The background map is produced by removing sources

detected in the wavelet detection, rescaling, and median smoothing. The

uncertainty in the background level should be less than 10%. The source

detection threshold is set to be lower than the final one, because the

later ML analysis will slightly revise the S/N ratios of individual sources.

This method is sensitive and has the well-defined detection sensitivity at

each point. But the method is still based on the image. Thus the source

centroid position is limited by the bin size of the image.

Maximum Likelihood analysis - A count-based analysis method for better

centroiding and count rate calculations. The sources detected in the map

detection are analyzed, using both the background map and a Gaussian PSF.

The accuracy of the position can be better than the pixel size. The

x and y uncertainties in the position are also calculated.

The final list of sources are merged and sorted. count rates individual bands are calculated with appropriate effective exposure corrections in these bands. These count rates can be used to estimate spectral properties. The total count rate and its uncertainty are also re-calculated, accordingly.

Only those above the detection threshold (e.g., S/N > 3 sigma) are listed.