Pedestal for NB required?

Elvie

Hello - first post. Just watched the video that describes the use of a pedestal to remove zero pixel values post dark subtraction.

After using pixel math to subtract the master dark from an uncalibrated sub I find no absolute zero pixels but many pixels have small values beginning in the fourth decimal place. When using pixel math again to replace any zero values with red I get a very red pixelated sub - pixel math does not seem to recognize such a small value and flags it as being zero.

I have calibrated the same sub with a small 100 pedestal addition and get a very nice image - a bit brighter, perhaps smoother. So since I have no truly zero pixels am I losing fidelity by not including a small pedestal value in my calibrations? And given that there are values so close to zero then maybe I should be using a pedestal just to be on the safe side?

For reference - this is with a cooled QHY268M for 300 second Ha subs under a bottle 4 sky.

Thanks for any thoughts.

BlockHead

Hi Elvie,

Even if you do not have an issue, a small pedestal of 100 counts is a small part of the full dynamic range it really shouldn't cause an issue... but...

As a matter of principle, your issue is an extra-ordinary one... so we need to go another step. In Pixel math... no matter how many digits to the right of the decimal point... anything that is there is NOT zero.

You are saying this is not true.

Could you generate that image with red pixels, and then with the readout preview, show it is not zero in a screenshot? (You should capture your Pixel Math expression as well.) If true, you have found a software bug that is really huge... so huge, it would have been noticed! 

-the Blockhead

Elvie

Hi Adam. 

Yes - no doubt I have made some very silly mistake and need to be shown the light. Here is the screen shot you requested. Tx.

BlockHead

Something is confusing me with your image. Your "RGB" image has RGB values that are unequal. When you convert a grayscale image to a color image (RGB)... the K value becomes 3 versions of K that are equal in each channel (it wouldn't be a gray image otherwise). That is not what you show!

I am pretty certain if we figure this out, we will understand what is going on. 

You are welcome to make your light frame and master dark available and I can look... but it is something to do with when you converted to RGB. 

-the Blockhead

Elvie

Hi Adam. The link below is to a shared dropbox folder: Ha sub and Master Dark. 

Captures were with NINA. The RGB conversion was via menu:  Image / Color Spaces

https://www.dropbox.com/sh/fkkf6soumugycbk/AAB7o3PRi-MCIArFqox5rpZOa?dl=0

Tx 

BlockHead

Everything works for me.

I fear there is something you did along the way... as you can see from my screen shots, the converted RGB image has the same values for each channel...and only zeros are lit up in red when I apply the pixel math expression. It is as if you applied something else to the image? Perhaps look at the history state and see?

-the Blockhead

Elvie

Thanks Adam - glad that it has worked for you - seriously.

I am at a loss to explain my problem. I will follow along wit the video again to see where I have lost my way.

Keep you posted.

Thanks again for taking the time too explore this.

Elvie

Hi Adam. In your 3rd capture above you show a zero value for a specific pixel location. I have done a simple PM light minus dark subtraction using the dropbox'ed images that I forwarded to you and I do not get a zero value for that specific pixel location; and in perusing the resulting image I see no zero valued pixels - please see attachment.

This is extremely odd. It is like I have a different math engine or incorrect settings somewhere.

BlockHead

Can you please  open up the "Destination" part of Pixel Math and take a screen shot?

-the Blockhead

Elvie

Sure - see attached.

FYI I calibrated some images with a pedestal of 100 and it removed 99.9% of the red pixels - so working as predicted despite my weird PM results. Also test integrated an Ha stack of 30 x 300s calibrated with pedestal 100 and using NSG for the first time - great script - together this gives a great result.

BlockHead

Hmmm...I had hoped you had "rescale" checked...that would have done the trick as to why the weird results. 

I guess I would literally need to watch you do your thing...I can't see how you get those values. 

You did get equal grayscale values last time..so that is good (one step closer). Now the zeros need to be there too.

At this point, I would literally make a recording of you doing this and showing you do not get the correct answer. The video could then be given to the PI developers as a bug.

-the Blockhead

Elvie

Hi Adam - as suggested - a recording of the steps. Would you mind reviewing the video - only a minute long - perhaps you will spot the error of my ways.

https://www.dropbox.com/s/qowav1zwfqxwxze/Pedestal.mov?dl=0

Tx for sticking with this.

BlockHead

Wow... I can't explain this.

I do see a tiny niggle that I cannot explain and I can't see how it was matter... but if you look at the name of your Master Dark- it appears you have opened it from the disk/harddrive. If you look at the title one part of the name ends in "51" ...and in the other part it just ends in "5."

It looks to me like it is correctly finding the zeros... for some reason the readout display isn't correct.

I am going to pass this on and see if I am missing something.

-the BLockhead

Elvie

Thanks for passing it on Adam. 

Pardon the niggle but I have been around the block (no pun intended) with this a few times and have tested things a few different ways. I have fixed the niggle with no effect on outcome. Funny (not really) part is I get the same result now as the first time I tested my dark subtraction.

Elvie

BlockHead

Juan did respond... he doesn't know either.

I think this is some kind of preview readout display issue (my guess).

I am giving him your files to look like. Likely... it will not reproduce. 

I do think at this point, if you ever install PI on another machine... you should try. 

-the Blockhead

BlockHead

Rob Pfile had an excellent idea..and I think it is the answer.

Probe size.

The readout preview has a probe size (number of pixels it is calculating values from). Mine is set to 1 pixel (which I think is the default). I *bet* yours is greater than one... this is why you never see zeros... you are looking at more than one pixel at a time!

Very subtle...I should have considered this.

-the Blockhead

Elvie

Probably quite system specific. Perhaps someone else will have a similar problem and a common thread will appear. Hopefully the root cause is not an issue that will surface in other ways to the detriment of any image processing.

Thanks for your time and effort.

Regards,

Elvie (aka Lenard)

 

BlockHead

No... you can change this... 

There is an arrow at the bottom of the screen. 

It has the readout parameters. Change the probe size to 1 pixel.

Then you and I will agree.

I really think this is the right answer.

Elvie

Success ==> probe size = 1

BIG thank you.