Changes between Version 5 and Version 6 of ppSub_vs_Hotpants

Timestamp:

Jan 21, 2010, 11:48:36 AM (16 years ago)

Author:

Paul Price

Comment:

Adding example of kernel scaling.

ppSub_vs_Hotpants

@@ -101 +101 @@
 ppSub (without adding photometry) has a running time about twice that of hotpants, but it is performing four times as many convolutions.  This better run time per convolution is probably due to ppSub doing convolutions via Fast Fourier Transforms, rather than directly.
 
-= Update =
+= Larger kernels =
 
 The following parameter variations result in much cleaner subtractions:
@@ -133 +133 @@
 35-36: 52
 }}}
+
+= Kernel scaling =
+
+Scaling of kernel basis sets has been implemented as of r26650 (in branches/eam_branches/20091201).  If SCALE in the PPSUB recipe is set to TRUE, then the ISIS kernel widths, the kernel size and stamp size are scaled by the maximum FWHM of the inputs, with maximum and minimum scaling values provided.  This results in fewer sources being detected in the subtracted images.
+
+The parameters used are:
+
+{{{
+KERNEL.SIZE: 45 (Half-size of kernel)
+STAMP.FOOTPRINT: 50 (Half-size of stamps)
+ISIS.WIDTHS: 2.9 4.8 8.2 13.8 23.3 (FWHM of Gaussians)
+ISIS.ORDERS: 4   2   2   2    2 (Polynomial order for Gaussians)
+SCALE: TRUE (Scale kernel parameters by the seeing)
+SCALE.REF: 12.0 (Reference for kernel parameter scaling)
+SCALE.MIN: 0.3 (Minimum scale value)
+SCALE.MAX: 1.2 (Maximum scale value)
+}}}
+
+Note that the stretch on the below image is from -25 to +25, rather than -100 to +100 as for the above images.
+
+[[Image(source:/trunk/doc/ppsub/pics/comparison/isis_radial_3.jpg)]]
+
+Cases with more than 10 sources detected in the subtracted image:
+
+{{{
+28-30: 54
+28-31: 43
+29-30: 48
+35-36: 81
+}}}