PSPHOT Kron Measurements Investigation

This page describes some investigations comparing the Kron radius and magnitudes measured by psphot

Overview

On April 12 Peter Draper pointed out that the Kron radii measured by psphot using the default recipe are much lower than those measured by SExtractor. He found that if the gaussian weighting function is removed, more sensible results are obtained. See plot at ​http://ps1sc.ifa.hawaii.edu/PS1wiki/index.php/KP12videocon_page#Kron_radii_-_possible_improvement

Using a set of 40 simulated images provided by Peter we have done many experiments with psphot to investigate the accuracy of the kron measurements. We reported April 30 at the DRAVG meeting that results closer to SEXtractor can be achieved with a prototype version of psphot that smooths the input image with a 1.7 pixel Gaussian prior to measuring the kron radius and flux. ​http://ps1sc.ifa.hawaii.edu/PS1wiki/index.php/DRAVG_telecon#Kron_radii_improvements

We have more carefully implemented the smoothing function and the code is now included in the ipp-trunk. There are several recipe values that now may be used to control the psphot kron measurements. Current defaault values are listed.

• PSPHOT_APPLY_WEIGHT T apply gaussian weight of sigma == windowRadius
• PSPHOT_APPLY_WINDOW T apply window function to downweight subtracted sources
• PSPHOT_KRON_SMOOTH F when true smooth the input image before measuring the kron radius
• PSPHOT_KRON_SIGMA 1.7 smoothing parameter
• PSPHOT KRON_NSIGMA 2 smoothing parameter

Upon reviewing the following results the differences between applying the weighting and window functions are not significant. We suggest turning them off. The smoothing yields kron radii closer to SExtractor with these simulated images. We need to investigate the performance impact on single frame psphot before deciding to turn it on.

Results

Kron Radius Histograms

The following images contain histograms of the measured Kron Radius for various recipe combinations.

• Green "Expected" Kron Radius for the simulated galaxy
• Grey Kron Radius measured by SExtractor (provided by Peter Draper)
• Red Kron Radius measured by psphot

Default recipe: Measure on smoothed image, gaussian weighting, window function

Window function no weighting

Weighting no window

No Window No weight

No window no weighting no smoothing

Difference In Kron Radius

Default Recipe

Window Function no weighting

Weighting, no window

No window, no weighting

No window, no weighting, no smoothing

Difference in Kron Magnitude

Default Recipe

Window Function no weighting

Weighting, no window

No window, no weighting

No window, no weighting, no smoothing

Statistics

Statistics for various recipe combinations
------------------------------------------

Fractional Difference in Kron Radius

( KR_SExtractor - KR_psphot ) / KR_Sextractor    All KR

Recipe                       mean       stdev      min     max         median      Npts
-------------------------------------------------------------------------------------------
default                  -0.201895    0.490706  -20.0476 0.924053    -0.103559     113705
nowin                    -0.2036380   0.689935 -165.129  0.909       -0.253494     113701
noweight                 -0.245435    0.83742  -129.597  0.900042    -0.143933     112929
noweight.nowin           -0.245726    0.834698 -129.597  0.948736    -0.226174     112930
nosmooth.noweight.nowin  -0.123882    0.830385 -101.242  0.921227     0.00175791   113087

Mean and meidan Kron raddii are closest to SExtractor with all of our various recipe choices off but
the standard deviation is lowest with our default recipe. This set of objects includes
small values of KR which psphot does not measure well (Minimum KR_psphot in this sampe is 7.1)



( KR_SExtractor - KR_psphot ) / KR_Sextractor   KR_SE > 10

Recipe                       mean       stdev      min       max      median      Npts
-------------------------------------------------------------------------------------------
default                 -0.0250245    0.22567    -4.66826  0.924053 -0.032233    42142
nowin                   -0.0257794    0.2263675  -4.576    0.909    -0.0291643   42140
noweight                -0.0402723    0.287004   -3.22148  0.900042 -0.0437856   42074
noweight.nowin          -0.0410538    0.287949   -4.44391  0.948736 -0.0435108   42078
nosmooth.noweight.nowin  0.0934499    0.53906  -101.242    0.921227  0.00175791  42138

Considering sources with SE kron radii above 10  the smallest stdev is also with the default recipe.
The window function appears to have has little effect on the statistics. The biggest improvement is provided
by the smoothing. Without smoothing the psphot KR is systematically low.

---------------------------------------------------------------------------------------------------------------
Kron Magnitude

KM_SExtractor - KM_psphot      All KR

Recipe                       mean       stdev      min    max   median     Npts
--------------------------------------------------------------------------------
default                   0.0144785  0.192932   -5.144   4.495  0.003226 113705
nowin                     0.0150963  0.1902736  -3.777   4.584  0.002172 113701
noweight                 -0.0860446  0.700536   -8.964   4.529 -0.004648 112929
noweight.nowin            0.0131     0.220402   -5.862   4.514  0.00044  112930
nosmooth.noweight.nowin  -0.014707   0.215786   -5.733   4.725 -0.03339  113087


KM_SExtractor - KM_psphot      KR_SE > 10 pixels

Recipe                       mean      stdev       min     max  median     Npts
--------------------------------------------------------------------------------
default                  -0.0229234  0.172541    -5.144  4.139 -0.001218  42142
nowin                    -0.0217864  0.1680391   -3.777  4.584 -0.0061890 42140
noweight                 -0.12206    0.674331    -7.844  3.337 -0.006119  42074
noweight.nowin           -0.0268161  0.187884    -4.832  4.514 -0.000118  42078
nosmooth.noweight.nowin  -0.065724   0.175386    -4.684  4.725 -0.035954  42138