Nicely done Frank! I just started adding to the pedestal during calibration. Thx for the real world examples. Really though a Pi earring? Gotta love your math...
Thank you very much for the very detailed and above all very interesting video on Pixelmath. The examples shown are very clearly explained. I have learned a lot again and will definitely test this in the next project. I found it very interesting how to find and count the zero values in the image.
@@setiv2 I'm very interested in something like that. I could watch it for two hours or more. I think the way you present your videos is great. Please keep up the good work.
Funnily enough my latest project would find your last point about combining two palettes in one image very useful. I had just started to wonder if it could be done...
Great video. I think one of the problems I have with PixelMath is not the mathematics behind it, but rather why you do it. For example in your signal boost, I don’t understand the logic of how the maths around the median removes the noise.
For that piece of pixelmath. If you just multiplied the whole channel you will multiply the background, blackpoint, shadows, noise floor, everything. By subtracting off the median first, everything less than the median (all that stuff above) is removed then you multiple. Once that is done add back in the median. A more elegant way would be to use an iif statement that if the pixel value is less than the median, return the pixel value, otherwise multiple the pixel value by some constant. I was trying to keep the first video more on the simple side to showing the range of things you can do
Hey Franklin! Your scripting is amazing! I have just watched you video regarding the statistical stretch video as well. Nevertheless I would like to comment on you Pixel Math video. I have saved a process icon with your formula to enhance the signal. After some try and error I came up with a slight enhancement of your formula that I would like to share with all. What I came up with is this Pixel Math: iif( $T>med($T), 1.5*($T-med($T))+med($T), ($T)) The math is checking for which pixels are above the median value, then applies your boost function that is multiplying the signal without the median and afterwards adds the median (noise) back in. If the pixels are equal or less the median the math just leaves these pixels as they are. The value 1.5 can be adjusted up or down depending on how much the signal should be boosted. What do you think about it?
Thanks for digging in and doing some trial and error pixel math! That is a perfect modification as without the iif statement the formula how i have it in the video actually darkens pixels below the median slightly. I want to make a second pixel math video with more complex functions. I didnt want to scare away any viewers on my first pixel math video. Quick word of caution on the signal boost formula. If you have pixel values greater than .667 it will start white clipping at 1.0.
@setiv2 Thank you very much for your response Franklin and the positive words! Not sure if you have tried my formula. I think that darkening the pixels that have median value make the appearance a bit unnatural. At least it has been the case on my trial photos. Nevertheless it can be working well depending on the quality of the photo and the object. Yeah! Clipping is a threat if someone does not have a good sense on numbers. The 1.5 multiplier can for sure be altered if needed. I need to dig into the operators to maybe find any that would only narrow the value to 1 but never reaches it. Splitting by channel and have the multiplier a variable symbol could also bring up some nice results. To be honest I am no mathematical expert I am just picking up ideas and try to enhance them for my needs. You are the genius behind the formula. 👍🏻
I have never used that setting in automatic. Maybe take one of your calibrated .xisf files from your narrowband data, like ha or sii. Run the exercise of looking for black clipped pixels, and see how well "automatic" worked
Very helpful video? I do have one question: Why do you apply the modified mask (where you add 0.5 to the inverted mask based on the bright parts of the galaxy) to the original, rather than the stars only, image? If I understand the formula for masks correctly, you are halving the brightness of the nebula, not the stars, when you apply it. What I can conclude from the PixInsight documentation makes it look a bit more complicated because processes are involved, but it still seems like having the mask on the stars only image makes more sense. Thanks.
No, the mask is halving the strength of the applied stars. You can think of it everywhere where the mask is white the full pixelmath equation is applied. Where it is half it is now a 50 50 blend of the original image (starless) and that of the final screened image.
@@setiv2 Thanks again! That must be what the documentation meant about the mask being applied to the processed version. But is there a reason to be giving equal weight to the starless and stars only images, at 50% each, in the area covered by the galaxy? I would think you’d want 50% of the stars only image and the original starless image combined using the usual ~((~A)*(~B)) formula.
No it doesnt show the stars only image. In the 50 50 area it would be a 50% blend of the starless and a 50% blend of the fully rescreened stars image. It is like you made a seperate full rescreened image and did a 50% blend with the starless image. That makes the stars 50% less prominent in the galaxy but the galaxy itself is completely unchanged
Nicely done Frank! I just started adding to the pedestal during calibration. Thx for the real world examples. Really though a Pi earring? Gotta love your math...
Huge math nerd here :)
Brilliant explanations. I might need to watch 2 or 3 more times when it comes to applying it in real life though!
Definitely worth doing the exercise of looking at clipped data in your NB calibrated images
Thank you very much for the very detailed and above all very interesting video on Pixelmath. The examples shown are very clearly explained. I have learned a lot again and will definitely test this in the next project. I found it very interesting how to find and count the zero values in the image.
I was worried people wouldnt make it that far on the video! Thanks for hanging on amd watching!
@@setiv2 I'm very interested in something like that. I could watch it for two hours or more. I think the way you present your videos is great. Please keep up the good work.
More PixelMath Please!!!! Great channel - you'll catch on!
This is fantastic content. So much of what we see on UA-cam is just click bait. But this is real gold. Thanks I’m enjoying the education.
great video! thanks for making it.
Excellent video, as always, Frank. Thanks for the tricks.
Great content your my goto source for learning Pixinsight. I love math can't wait to try some of this out.
Funnily enough my latest project would find your last point about combining two palettes in one image very useful. I had just started to wonder if it could be done...
More pixel math(s) please!
It is math not maths in this country hahaha. Good to see people want more pixelmath though.
👍 tes vidéos…merci à toi.
Merci!
excellent, thanks
Great video. I think one of the problems I have with PixelMath is not the mathematics behind it, but rather why you do it. For example in your signal boost, I don’t understand the logic of how the maths around the median removes the noise.
For that piece of pixelmath. If you just multiplied the whole channel you will multiply the background, blackpoint, shadows, noise floor, everything. By subtracting off the median first, everything less than the median (all that stuff above) is removed then you multiple. Once that is done add back in the median. A more elegant way would be to use an iif statement that if the pixel value is less than the median, return the pixel value, otherwise multiple the pixel value by some constant. I was trying to keep the first video more on the simple side to showing the range of things you can do
Excellent! Thank you.
Really helpful tutorial Thanks
Nickel 👌…merci encore c’est cool 👍…!.
Hey Franklin! Your scripting is amazing! I have just watched you video regarding the statistical stretch video as well. Nevertheless I would like to comment on you Pixel Math video. I have saved a process icon with your formula to enhance the signal. After some try and error I came up with a slight enhancement of your formula that I would like to share with all.
What I came up with is this Pixel Math: iif( $T>med($T), 1.5*($T-med($T))+med($T), ($T))
The math is checking for which pixels are above the median value, then applies your boost function that is multiplying the signal without the median and afterwards adds the median (noise) back in. If the pixels are equal or less the median the math just leaves these pixels as they are.
The value 1.5 can be adjusted up or down depending on how much the signal should be boosted.
What do you think about it?
Thanks for digging in and doing some trial and error pixel math! That is a perfect modification as without the iif statement the formula how i have it in the video actually darkens pixels below the median slightly. I want to make a second pixel math video with more complex functions. I didnt want to scare away any viewers on my first pixel math video.
Quick word of caution on the signal boost formula. If you have pixel values greater than .667 it will start white clipping at 1.0.
@setiv2 Thank you very much for your response Franklin and the positive words! Not sure if you have tried my formula. I think that darkening the pixels that have median value make the appearance a bit unnatural. At least it has been the case on my trial photos. Nevertheless it can be working well depending on the quality of the photo and the object.
Yeah! Clipping is a threat if someone does not have a good sense on numbers. The 1.5 multiplier can for sure be altered if needed. I need to dig into the operators to maybe find any that would only narrow the value to 1 but never reaches it. Splitting by channel and have the multiplier a variable symbol could also bring up some nice results.
To be honest I am no mathematical expert I am just picking up ideas and try to enhance them for my needs. You are the genius behind the formula. 👍🏻
Great tutorial, thank you!
Interesting.
Great video. Thanks. I use pedestal in "automatic" in WBPP. Is it better to choose those 600 or 800 values?
I have never used that setting in automatic. Maybe take one of your calibrated .xisf files from your narrowband data, like ha or sii. Run the exercise of looking for black clipped pixels, and see how well "automatic" worked
Very helpful video? I do have one question: Why do you apply the modified mask (where you add 0.5 to the inverted mask based on the bright parts of the galaxy) to the original, rather than the stars only, image? If I understand the formula for masks correctly, you are halving the brightness of the nebula, not the stars, when you apply it. What I can conclude from the PixInsight documentation makes it look a bit more complicated because processes are involved, but it still seems like having the mask on the stars only image makes more sense. Thanks.
No, the mask is halving the strength of the applied stars. You can think of it everywhere where the mask is white the full pixelmath equation is applied. Where it is half it is now a 50 50 blend of the original image (starless) and that of the final screened image.
@@setiv2 Thanks again! That must be what the documentation meant about the mask being applied to the processed version. But is there a reason to be giving equal weight to the starless and stars only images, at 50% each, in the area covered by the galaxy? I would think you’d want 50% of the stars only image and the original starless image combined using the usual ~((~A)*(~B)) formula.
No it doesnt show the stars only image. In the 50 50 area it would be a 50% blend of the starless and a 50% blend of the fully rescreened stars image. It is like you made a seperate full rescreened image and did a 50% blend with the starless image. That makes the stars 50% less prominent in the galaxy but the galaxy itself is completely unchanged
@@setiv2 Thanks for the quick reply. I’ll have to think about how that works.
No worries. Mathematically in the 50 50 area the final image is 0.5*A+0.5*~(~A×~B) assuming A is ur starless and B is your stars only
Yeah nah I hate Pixelmath, prefer the scripts created by those in the know.
Scripts can be so one off. Plus programming in pixinsight java is more of a pain than pixelmath lol