Changeset 30220

Timestamp:

Jan 7, 2011, 2:09:48 PM (15 years ago)

Author:

eugene

Message:

the culling process was slightly wrong: we should keep any significant peak which is separated from a brighter peak by the min threshold (nominally 1 sigma above sky); the old code was rejecting things if the min threshold was higher than the downward deviation (true for faint sources)

File:

• branches/eam_branches/ipp-20101205/psModules/src/objects/pmFootprintCullPeaks.c (modified) (3 diffs)

branches/eam_branches/ipp-20101205/psModules/src/objects/pmFootprintCullPeaks.c

@@ -28 +28 @@
   * Examine the peaks in a pmFootprint, and throw away the ones that are not sufficiently
   * isolated.  More precisely, for each peak find the highest coll that you'd have to traverse
-  * to reach a still higher peak --- and if that coll's more than nsigma DN below your
+  * to reach a still higher peak --- and if that coll's more (less?) than nsigma DN below your
   * starting point, discard the peak.
   */
@@ -99 +99 @@
         float threshold = flux - nsigma_delta*stdev_pad;
 
-        if (isnan(threshold) || threshold < min_threshold) {
+        if (isnan(threshold)) {
             // min_threshold is assumed to be below the detection threshold,
             // so all the peaks are pmFootprint, and this isn't the brightest
@@ -109 +109 @@
             threshold = subImg->data.F32[y][x] - 10*FLT_EPSILON;
         }
+
+        if (threshold < min_threshold) {
+            threshold = min_threshold;
+        }
 
         // init peakFootprint here?