Data Rejection Basics

Document Sample
Data Rejection Basics Powered By Docstoc
					CCDStack Take 2
 CCDStack Version 2 Tutorial
      B. Waddington
        4/19/2011
             Stack-based Rejection

Compute
statistics on
each “column”
of pixels in a
stack of
images…
              Stack-based Rejection
   Basic Concept
       Estimate the “most likely value” for each column in the
        stack – typically the “average” value
       Look at the dispersion of values in that column relative to
        the average
       Reject those pixels that look “too far off the mark”
   Rejection occurs for individual pixels in each
    image, not the whole image
   Most of the data rejection techniques are all about
    defining “too far off the mark”           │
              Stack-based Rejection

Individual pixels
in each column
are rejected as
being outliers…
Rejected Pixels
   What happens to the pixels that are
    rejected?
   They are assigned a “rejected value” and
    are excluded from subsequent calculations
    – but the original values are not lost
   They may need to be given “imputed”
    values if there are no “good” values to use
    in their place (e.g. blooms)
Data Rejection Choices – Sigma Clip
   Computes the average and std. deviation of
    the pixel stack; keeps only those pixels that
    lie within a sigma factor of the average
   Strong noise rejection but needs 10+
    images for best results
   Allows “tuning” based on the sigma factor
    you specify
   Can be iterated                      │
Sample Values in a “Pixel Stack”
                   Sampled Values of Single Pixel Position

                  12
                  10                                                Bogus
      Frequency




                   8                                                values
                   6
                   4
                   2
                   0




                                                                             e
                    00

                           50

                                  00

                                         50

                                                00

                                                       50

                                                              00

                                                                     50

                                                                          or
                  22

                         22

                                23

                                       23

                                              24

                                                     24

                                                            25

                                                                   25

                                                                          M
                                              Pixel Value
Rejection by Sigma Clip
                   Sampled Values of Single Pixel Position
                                                        3σ
                 12                                2σ
                 10
     Frequency




                  8                                                      3-Sigma
                  6                                                      Clip
                  4
                  2
                  0




                                                                              e
                   00

                          50

                                 00

                                        50

                                               00

                                                      50

                                                             00

                                                                    50

                                                                           or
                 22

                        22

                               23

                                      23

                                             24

                                                    24

                                                           25

                                                                  25

                                                                         M
                                             Pixel Value
Rejection by Sigma Clip
                   Sampled Values of Single Pixel Position
                                                        3σ
                 12                                2σ
                 10
     Frequency




                  8
                                                                   2-Sigma
                  6
                  4
                      2-Sigma                                      Clip
                  2
                  0




                                                                             e
                   00

                          50

                                 00

                                        50

                                               00

                                                      50

                                                             00

                                                                    50

                                                                          or
                 22

                        22

                               23

                                      23

                                             24

                                                    24

                                                           25

                                                                  25

                                                                         M
                                             Pixel Value
Poisson Data Reject
   Uses a theoretical “spread” (sigma)
    assuming a Poisson distribution of data
   Poisson sigma == Sqrt(average signal)
       Average signal value calculation includes use of
        camera gain and read noise
       Camera values need to be correct
Poisson Data Reject
   Can be used on small stacks – even two or
    three images
   Very strong noise rejection
   May damage stars if the star sizes are not
    pretty similar across frames
   You need to try various sigma factors and
    look closely at the results
Workflow
   Calibration
   Registration
   Normalization
   Data rejection
   Combine (mean)
   DDP processing for luminance
   Sharpening for “preview”
Registration Tips
   Pick whatever image is well-framed as reference
   CCDIS is preferred – requires CCDInspector plug-
    in ($90)
   You can mix and match alignment methods as
    desired
   Help file has good explanations
   “Apply” alternatives
       Nearest neighbor – image must be well-sampled,
        alignment must be dead-on; be SURE you blink the stack
       Quadratic B-Spline is a safer choice
Normalization
   Pick your best quality image in the stack
   Normalization is CRITICAL to data rejection
   It can be repeated multiple times if needed
Data Rejection Tips
   Use your best-quality image
   Try for 1-2% data rejection on this
    reference image by adjusting the “factor”
    and looking at the “Info” window
   STD Sigma Reject is a good all-purpose
    choice
   Poisson Reject is well-suited for small
    stacks or nasty problems
Color Combination Tips
   Create an LRGB composite
   Use the background adjustment option
   Boost the contrast for better visualization
   White adjustment options
       Choose a known-white object
       Select an entire spiral galaxy
   Very good starting point on color balance
CCDStack Help
   Version 2 includes much-improved help
       Context-sensitive help (F1)
       Expanded information on various processing
        options
   CCDWare support forums are monitored
    and used by program authors
References
   Streaming video tutorial:
    http://www.ccdware.com/tutorials/ccdstack/overview.html
   Version 1 tutorial:
    http://www.ccdware.com/Files/CCDStack%20Tutorial.pdf
   Original SIG presentation:
    http://www.oc-aisig.org/docs/tutorials/


   Where to get CCDStack:
    http://www.ccdware.com/

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:13
posted:5/31/2012
language:English
pages:18