Author Topic: PRE PROCESSING IN PI VERSUS BPP OR DSS IS FAILING WITH USE OF ASTRONOMIK CLS  (Read 2090 times)

Offline ballyhoo

  • Member
  • *
  • Posts: 74
I really do not know what setting could have derailed me, however using the full PI soup to nut pre-processing algorithm is supposed to be an asset, rather than the opposite.    In this integration of The Swan I have 13 x 180 subs, 16 flats, 15 darks and about 100 bias frames. I used an Astronomics CLS filter since I integrated from my front door and there is light pollution. I follow the PI flow  per Keller, starting with calibration of flat frames, etc,integration of flats, CC, DEB, and SA.  below is what I am left with after ABE or DBE. Now, the link I provide on astrobin is a BPP version -- much nicer.  I have no idea where to explore what went wrong w the PI pre-process procedures. I also tried DSS and that came out fine too.

Here is my Fit file.
https://www.dropbox.com/s/pbx494g39e07civ/integration_DBE.fit?dl=0


BPP version:

https://www.astrobin.com/416797/B/?nc=user

THANK YOU FOR ANY INSIGHT YOU MAY HAVE!

Offline STEVE333

  • PixInsight Addict
  • ***
  • Posts: 231
    • sk-images
I downloaded your fit file and separated it into the R/G/B channels. The G and B channels look normal. However, there is definitely something wrong with the R channel, and, that is what is causing your image to be all greenish.

Steve
Telescopes:  WO Star71 ii, ES ED102 CF
Camera:  Canon T3 (modified)
Filters:  IDAS LPS-D1, Triad Tri-Band, STC Duo-Narrowband
Mount:  CEM40 EC
Software:  BYEOS, PHD2, PixInsight

http://www.SteveKing.Pictures/

Offline bulrichl

  • PixInsight Guru
  • ****
  • Posts: 524
I appended the Histogram of the RGB image and the image of the extracted R channel (integration_DBE.JPG). It looks like a mask was applied to the R channel.

I guess there is something wrong with the input for ImageIntegration, to be specific: with one or several of the target images. You will have to take a look at the images that you integrated with ImageIntegration, e.g. with the Blink process. Identify the culprit and repeat the ImageIntegration process omitting or exchanging the target images in question.

Bernd

Offline acmalko

  • Member
  • *
  • Posts: 54
Hi,

show us your masterflat...

Offline ballyhoo

  • Member
  • *
  • Posts: 74
Thanks all, am investigating...

here is the master flat:

https://www.dropbox.com/s/b1bbj00ltukat43/integration%20flat.fit?dl=0


EDIT

I SEARCHED IN BLINK. NOTHING DIFFERENT IN ANY OF THE FRAMES.
« Last Edit: 2019 July 25 09:16:12 by ballyhoo »

Offline acmalko

  • Member
  • *
  • Posts: 54
Ok, bad masterflat, saturated.

You can presume it by differents ways.
1- problem of color on your stack
2- you still see dust after pretreatment

Offline ballyhoo

  • Member
  • *
  • Posts: 74
Thank you,, Does this mean I over exposed the flat?  I hav an ASI 294 color and I think the E value is 60k so I exposed to 29000

If the flat is the reason why the integration looks the way it does, why does the integration come out okay in BPP And DSS?


Offline acmalko

  • Member
  • *
  • Posts: 54
Flats over exposed, yes, that's exactly what I mean. In your case, the green and blue channels are much more high than the red one, certainly saturated, overexposed. You can try without the masterflat, I'm pretty sure you'll have a better result of your stack.
Why is it ok with DSS ? I don't know
How much expose your flat ? I don't know actually. Mines are done easily with a canon camera and I use a pink screen to be sure that the 3 channels are in the same area of an histogram


Offline ballyhoo

  • Member
  • *
  • Posts: 74
Well, you get the cake today! (This bloke is a genius!)


I went through the entire process without flats, and this is what I got:

(I am at 29K ADU w flats. How far down should I go? 20k?)


EDIT:

VIVA LA FRANCE!!    :)
« Last Edit: 2019 July 25 10:27:13 by ballyhoo »

Offline bulrichl

  • PixInsight Guru
  • ****
  • Posts: 524
Even though the issue seems to be caused by the MasterFlat, this is certainly not the case. Something different is going on here.

The MasterFlat is definitely not saturated. Statistics tool says maximum = 0.574. After a split into CFA channels, the results are (in the Statistics process 'Unclipped' is disabled):

Code: [Select]
    CFA0 CFA1     CFA2 CFA3
count (%)   100.00000           100.00000           100.00000           100.00000
count (px)  2923592             2923592             2923592             2923592
mean        1.484427610e-01     5.348702580e-01     5.349518453e-01     5.358891165e-01
median      1.494930387e-01     5.369691252e-01     5.370509623e-01     5.383101701e-01
stdDev      5.705807990e-03     1.809740888e-02     1.808897539e-02     1.673962951e-02
avgDev      5.592659166e-03     1.778350117e-02     1.778313594e-02     1.640620176e-02
MAD         5.509970383e-03     1.807649392e-02     1.808603787e-02     1.640082490e-02
minimum     5.841452257e-02     4.034291209e-01     4.052959682e-01     4.009656609e-01
maximum     1.614647359e-01     5.729865431e-01     5.740009545e-01     5.712071060e-01

The CFA1, CFA2 and CFA3 channels are very close together, but the CFA0 channel (red) is weak. This is due to the light source and the high sensitivity of the sensor for green light.
The strongly different illumination of green and blue versus red may be not optimal, but in my view, the MasterFlat is not the cause of the problem.

(Besides, your flats were calibrated with the setting 'No normalization' in the section 'Pixel Rejection (1)'. Recommended is 'Equalize fluxes'. - However, this will not be the reason for your issue as well.)

Does the integration of light frames (that is: BEFORE applying DBE) also show this artifact?

If yes, repeat the ImageIntegration step with the previously checked input files (fully calibrated light frames).

I also have an ASI294 and in my MasterFlats the red channel is also weak, albeit not as extreme as in your MF. I never saw a problem like this before.

Bernd

Offline acmalko

  • Member
  • *
  • Posts: 54
Thanks for your explanations

A question to help me understand: had you 29.000 ADU on each channel ?
Secondly: why such high value ? (I know the theory, this is a serious question ;) )

Offline acmalko

  • Member
  • *
  • Posts: 54
Oh and I think the statistics and histogram processes are in 16bits but you use a 14bits camera so personnally, I multiply by 4 the statistics numbers

Offline dave_galera

  • PixInsight Addict
  • ***
  • Posts: 261
    • QDigital Astro
I know this won't help with the problem if its not the master flats that are the problem, but may be useful with determining the correct ADU value for them:

http://www.telescopesupportsystems.com/thrushobservatory.org/Tips/Digital%20Imaging/flatfieldcalc.htm

BTW I aim for a value between 20K to 22K

Dave
Dave

Offline bulrichl

  • PixInsight Guru
  • ****
  • Posts: 524
In this outputs of Statistics, the output format was set to 'Normalized Real [0,1]'. By multiplying these values by 65536, the 16-bit [0,65535] representation ar obtained.

This means that the CFA channels have following means (and in parentheses: the medians):
Code: [Select]
CFA0:  9728  (9797)
CFA1: 35053 (35190)
CFA2: 35058 (35196)
CFA3: 35119 (35278)

The ADC of the ASI294 has only 14 bit, but the data of the ASI294 are multiplied by factor 4 to the range 0 to 65535 by the camera driver. The values above for the green and blue channels are therefore OK.

Bernd

Offline ballyhoo

  • Member
  • *
  • Posts: 74
To be clear, no such artifact is visible in the light frames.  And if it is not an issue with the flats, why then removing them from the integration removes the artifact?

Not sure how to approach this issue, other than when I use my Astronomik filter to process in BPP.

Uh-ooh, I just bought the Triad filter.  I wonder what awaits me with that filter.