Bayer/X-Trans Drizzle Issue with OSC Data (2.0.0-beta15 & 16)

I’ve run into a drizzle issue in 2.0.0-beta15.  The issue did not appear to be resolved in the beta16 version.

I’m on a Dell Windows 10 workstation (Version 10.0.19044 Build 19044) with 32 GB RAM, and an Intel(R) Xeon(R) CPU E3-1240 v5 @ 3.50GHz, 3504 Mhz, 4 Core(s), 8 Logical Processor(s).  Graphics Adapter is NVIDIA Quadro M4000.

All data were captured using a Canon R5 OSC body.  For the test cases I loaded all calibration frames (flats, darks, and bias) anew.  There were only 10 lights so the case would run faster.  My data are undersampled, but were not intentionally dithered on the EQ mount I used.  Nevertheless, I think this is a case of something wrong in the code (in concert with my subpar data).

Unless otherwise indicated, all APP processing settings were default:

1) LOAD Tab

I unchecked the top 3 checkboxes

I ran 3 test cases.  The first and last produced the expected result.  The second test case (using Bayer/X-Trans Drizzle) did not produce the expected result.

The first test case used all default settings (i.e. no drizzle).  The result appeared to be normal to my eye.  I didn’t bother including that image as it was just a typical, light-polluted result that looked as expected.

The second test case changed the 6) INTEGRATE Tab INTEGRATE mode to Bayer/X-Trans drizzle.  This result did not look normal.  There was red and green shading (see below).

Below is a 500% zoom that may help in your assessment.  That same zoom level on the other 2 images in this test looked as expected.  They did not have the red block artifacts.

For reference, I had previously run the full data set of 119 lights, integrating a crop of the Reference Frame, and the red appeared as vertical bands that went through much of the image; the oddities were far more pronounced (see below). 

Here is the .fits header data from APP for the second case which exhibited the issue (for reference).  Note that I provided the header for the beta15 version.  The beta16 software did not appear to resolve the issue.

FITS HDUs: 1

HDU1 - SIMPLE  =                    T / Java FITS: Tue Apr 11 14:04:23 EDT 2023       

HDU1 - BITPIX  =                  -32 / bits per data value                           

HDU1 - NAXIS   =                    3 / number of axes                                 

HDU1 - NAXIS1  =                 8200 / size of the n'th axis                         

HDU1 - NAXIS2  =                 5482 / size of the n'th axis                         

HDU1 - NAXIS3  =                    3 / size of the n'th axis                          

HDU1 - EXTEND  =                    T / Extensions are permitted                      

HDU1 - BSCALE  =                  1.0 / scale factor                                  

HDU1 - BZERO   =                 0.E0 / no offset                                      

HDU1 - DATE    = '2023-04-11T18:23:12' / creation date of Integration                 

HDU1 - SOFTWARE= 'Astro Pixel Processor by Aries Productions' / software              

HDU1 - VERSION = '2.0.0-beta15'       / Astro Pixel Processor version                 

HDU1 - INTEGRAT= 'Integration'        / integration of light frames                   

HDU1 - CFAIMAGE= 'no      '           / Color Filter Array pattern                    

HDU1 - NOTE-1  = 'INTEGRATION METADATA'                                                

HDU1 - EXPTIME =                 9.E2 / exposure time (s)                             

HDU1 - NUMFRAME=                   10 / number of frames used in this integration     

HDU1 - BG-1    = ' 5.2125E-02'        / background estimate of channel 1              

HDU1 - BG-2    = ' 1.3887E-01'        / background estimate of channel 2               

HDU1 - BG-3    = ' 9.0407E-02'        / background estimate of channel 3              

HDU1 - SCALE-1 = ' 2.9664E-02'        / dispersion of channel 1                       

HDU1 - SCALE-2 = ' 1.0033E-02'        / dispersion of channel 2                        

HDU1 - SCALE-3 = ' 7.4355E-03'        / dispersion of channel 3                       

HDU1 - NOISE-1 = ' 2.8356E-02'        / noise level of channel 1                      

HDU1 - NOISE-2 = ' 2.8735E-03'        / noise level of channel 2                      

HDU1 - NOISE-3 = ' 3.0756E-03'        / noise level of channel 3                      

HDU1 - SNR-1   = ' 9.7136E-01'        / Signal to Noise Ratio of channel 1            

HDU1 - SNR-2   = ' 4.2730E+00'        / Signal to Noise Ratio of channel 2            

HDU1 - SNR-3   = ' 2.7735E+00'        / Signal to Noise Ratio of channel 3            

HDU1 - NOTE-2  = 'NR = Noise Reduction'                                               

HDU1 - NOTE-3  = 'medNR = noise in median frame / noise in integration'               

HDU1 - NOTE-4  = 'refNR = noise in reference frame / noise in integration'            

HDU1 - NOTE-5  = 'ideal noise reduction = square root of number of frames'            

HDU1 - NOTE-6  = 'the realized/ideal noise reduction ratio should approach 1 ideally' 

HDU1 - NOTE-7  = 'the effective noise reduction has a correction for'                 

HDU1 - NOTE-8  = 'dispersion change between the frame and the integration'            

HDU1 - NOTE-9  = 'because dispersion and noise are correlated'                        

HDU1 - MEDNR-1 = ' 1.4185E-01'        / median noise reduction, channel 1             

HDU1 - MEDNR-2 = ' 1.5745E+00'        / median noise reduction, channel 2             

HDU1 - MEDNR-3 = ' 1.0722E+00'        / median noise reduction, channel 3             

HDU1 - REFNR-1 = ' 1.4134E-01'        / reference noise reduction, channel 1          

HDU1 - REFNR-2 = ' 1.5833E+00'        / reference noise reduction, channel 2          

HDU1 - REFNR-3 = ' 1.0631E+00'        / reference noise reduction, channel 3          

HDU1 - IDNR-1  = ' 3.1623E+00'        / ideal noise reduction, channel 1              

HDU1 - IDNR-2  = ' 3.1623E+00'        / ideal noise reduction, channel 2              

HDU1 - IDNR-3  = ' 3.1623E+00'        / ideal noise reduction, channel 3              

HDU1 - RATNR-1 = ' 4.4855E-02'        / realized/ideal noise reduction ratio, channel 1

HDU1 - RATNR-2 = ' 4.9789E-01'        / realized/ideal noise reduction ratio, channel 2

HDU1 - RATNR-3 = ' 3.3905E-01'        / realized/ideal noise reduction ratio, channel 3

HDU1 - MEDENR-1= ' 6.2124E-01'        / effective median noise reduction, channel 1   

HDU1 - MEDENR-2= ' 3.4736E+00'        / effective median noise reduction, channel 2   

HDU1 - MEDENR-3= ' 1.0298E+00'        / effective median noise reduction, channel 3   

HDU1 - REFENR-1= ' 6.1540E-01'        / effective reference noise reduction, channel 1

HDU1 - REFENR-2= ' 3.5016E+00'        / effective reference noise reduction, channel 2

HDU1 - REFENR-3= ' 1.0135E+00'        / effective reference noise reduction, channel 3

HDU1 - NORMMODE= 'regular '           / normalization mode                            

HDU1 - NORMMETH= 'multiply-scale'     / normalization method                          

HDU1 - NORMSCAL= 'BWMV    '           / normalization scale/dispersion calculation    

HDU1 - NORM-BGN= 'neutralize bg'      / normalization background neutralization       

HDU1 - NOTE-10 = 'REFERENC tag: used reference frame'                                 

HDU1 - REFERENC= '_87A9843.CR3'                                                       

HDU1 - COMPMODE= 'full    '           / composition mode                              

HDU1 - REGMODE = 'normal  '           / registration mode                             

HDU1 - REGMODEL= 'projective'         / registration model                            

HDU1 - OPT-DC  = 'disabled'           / optical distortion correction                 

HDU1 - WEIGHTS = 'equal   '           / integration weights                           

HDU1 - INT-METH= 'average '           / integration method                            

HDU1 - OUTL-REJ= 'winsorized rejection' / outlier rejection filter                    

HDU1 - OUTL-DP = '6       '           / outlier rejection diffraction protection      

HDU1 - OUTL-KL =                 6.E0 / outlier rejection kappa low                   

HDU1 - OUTL-KH =                2.1E0 / outlier rejection kappa high                  

HDU1 - INT-MODE= 'Bayer/X-Trans drizzle' / integration mode                           

HDU1 - DRZ-DROP=                 1.E0 / drizzle droplet size                          

HDU1 - DRZ-KERN= 'topHatKernel'       / drizzle kernel                                

HDU1 - INTSCALE=                 1.E0 / integrate scale                                

HDU1 - NOTE-11 = 'PROJECT tag: projection type'                                       

HDU1 - PROJECT = 'rectilinearProjection'                                              

HDU1 - MBBLEND = 'no MBB  '           / multi-band blending                            

HDU1 - LNC-DEG = 'noLNC   '           / Local Normalization Correction not applied    

HDU1 - AD-PED  =                 0.E0 / adaptive pedestal from data calibration       

HDU1 - END                                                                             

The third test case changed the 6) INTEGRATE Tab INTEGRATE mode to drizzle.  I selected No when warned I should be using Bayer/X-Trans drizzle, and forced APP to use regular drizzle.  The result appeared to be normal to my eye.  Again, I didn’t bother including that image as it was just a typical, light-polluted result that looked as expected.

So in summary, 1 form of drizzle seemed to work fine, but one did not.

Please let me know if I should upload any files (such as the .fits result).

Note that I sent an email to the support email address a few days back but have not seen a response.  I'm guessing the forum is the preferred approach for these types of issues.  If otherwise, just let me know.

Thanks.