Changeset 26703 for branches/eam_branches/20091201/ippconfig

Timestamp:

Jan 28, 2010, 4:16:19 AM (17 years ago)

Author:

Paul Price

Message:

Solve for the normalisation of the convolution kernel completely separately from that of the other kernel terms. We use a 'normalisation window', basically an aperture in which we measure the flux of the stamps to determine the normalisation value. This introduces a new recipe parameter, NORM.FRAC, which is the fraction of the total in the window at which we set the normalisation window/aperture. The quality of the result seems to be rather sensitive to this value. Also found a bug in the imposition of the penalty function in dual convolution mode: it wasn't being set for the second solution part. This seems to produce decent dual convolution subtractions.

File:

• branches/eam_branches/20091201/ippconfig/recipes/ppSub.config (modified) (2 diffs)

branches/eam_branches/20091201/ippconfig/recipes/ppSub.config

@@ -2 +2 @@
 
 KERNEL.TYPE     STR     ISIS_RADIAL     # Kernel type to use (POIS|ISIS|SPAM|FRIES|GUNK|RINGS)
-KERNEL.SIZE     S32     30              # Kernel half-size (pixels)
+KERNEL.SIZE     S32     35              # Kernel half-size (pixels)
 SPATIAL.ORDER   S32     1               # Spatial polynomial order
 REGION.SIZE     F32     0               # Iso-kernel region size (pixels)
 SOURCE.RADIUS   F32     3.0             # Source matching radius (pixels)
 STAMP.SPACING   F32     300             # Typical spacing between stamps (pixels)
-STAMP.FOOTPRINT S32     35              # Size of stamps (pixels)
+STAMP.FOOTPRINT S32     40              # Size of stamps (pixels)
 STAMP.THRESHOLD F32     5               # Flux threshold for stamps (stdev above background)
 STRIDE          S32     100             # Size of convolution patches (pixels)
-ITER            S32     10              # Number of rejection iterations
+ITER            S32     2               # Number of rejection iterations
 REJ             F32     2.5             # Rejection level (std dev)
 KERNEL.ERR      F32     0.05            # Relative systematic error in kernel
 SYS.ERR         F32     0.1             # Relative systematic error in images
 SKY.ERR         F32     0.0             # Relative systematic error in images
+NORM.FRAC       F32     0.03            # Fraction of window for normalisation window
 
 MASK.VAL        STR     MASK.VALUE,CONV.BAD     # Mask value for input
@@ -73 +74 @@
 STACK   METADATA
         KERNEL.TYPE     STR     ISIS    # Kernel type
+        SCALE.REF       F32     9.0     # Reference for kernel parameter scaling
 END