Changeset 42088 for trunk/psphot/src/psphotSignificanceImage.c

Timestamp:

Feb 28, 2022, 2:40:32 PM (4 years ago)

Author:

eugene

Message:

for bright stars in stacks, the peak in the S/N image can have a central divot if the variance of the central pixels is relatively large compared to the ring of pixels further from the center. the detected peaks are then not well centered. this can be avoided if we detect on the signal image itself (after first determining an appropriate flux value for the desired S/N limit); added an option to use the signal image for the first pass peak detetion step

Location:

trunk/psphot

Files:

• . (modified) (1 prop)
• src/psphotSignificanceImage.c (modified) (2 diffs)

trunk/psphot/src/psphotSignificanceImage.c

@@ -105 +105 @@
               //              float v2 = value + PS_SQR(value/100.0);
               // CZW 2013-06-20: I don't think this hack was helping.
+              // EAM 2021-12-21: see note below about divots in the significance image
               float v2 = value;
               smooth_wt->data.F32[j][i] = factor * PS_SQR(v2) / smooth_wt->data.F32[j][i];
@@ -122 +123 @@
     }
     psImageConvolveSetThreads(oldThreads);
+
+    // We now have the significance image and the signal image.  In some cases (e.g.,
+    // stacks), the variance on pixels in the cores of stars is elevated compared to pure
+    // poisson statistics.  In this case, especially at high signal levels, the ratio of
+    // signal / noise in the core of the star can be lower than the surrounding ring of
+    // pixels.  This results in a divot in the center of the star in the significance
+    // image.  the apparent peak of the significance is then not centered on the star and
+    // chaos ensues.  A possible fix is to use the signal image for signficance for
+    // the high S/N detection pass.
 
     pmReadout *significanceRO = pmReadoutAlloc(NULL);