Great new tool. It didn’t work on my RGB images captured with an OSC camera. Should they be captured via a monochrome camera or do I need to split the channels for it to work?
I am amazed at the quality of some of the free options/processes available now. But I wouldn't expect "default" values to generally give you a definitive evaluation of "better" and "worse". I assume the sliders are there for a reason. For example, if you moved the detail slider on NXT, would you get a "better" result?
I agree with you there. The sliders do work to an extent in NoiseXterminator but create other issues. I’m in process of a video going over that as well as go more in depth with deepsnr.
Great vid. Can't wait to try Deep snr and compare to NoiseX I also need to look into balancing color chanels. I have never done so. Is it used for osc cams too?
Thank you! Yes, color balancing channels is for OSC as well. I also just realized I didn’t put a link to my workflow video 🤣, thank you for the question because I just realized that! Check out this video and you’ll see exactly how to do it :) th-cam.com/video/LS3Gpx75hhU/w-d-xo.htmlsi=em3u6yPe4221kUIm
Great question! Yes, I’m using a OSC camera for this example. It’s a little bit out of bounds of the software’s intention and also accounts for the quirk of the behavior causing it to need to be run early on. I’m in progress of a follow up video going more in depth with it. I did see that it’s not working for you. Can you describe the workflow that you’re using from start to the point of running it?
Looks very promising. I hope you can do OSC also in the future as splitting rgb channels is not in my workflow. I just tried DeepSNR on a OSC file and it doesn’t work.
I am very excited to see where this process goes! Do me a favor. Try running DeepSNR at the very start of your workflow as the first item and let me know if this allows it to work. The image in working on in this video is a OSC
Just to confirm, you copied both files from the DeepSNR download and pasted in the PixInsight BIN folder? Also, how do you do your stacking and what file type is it?
@@Hidden.Light.Photography Yup, it works but don’t like the result. I stack with Siril and use 32bit fits files. I don’t like what it does to the background. It takes noise and makes little stars of it.
That’s not good. It might be the way Siril is stacking and preprocessing the image. Unfortunately this may be one that needs to be fully developed before using. There are some quirks using my OSC data as well which is done through WBPP in PixInsight that cause me to have to run DeepSNR in the very beginning for best results. I do truly believe it’s worth the wait for this process seeing the absolute potential already.
Just to make sure, you downloaded the correct PixInsight version? If so, the software is still being developed, so I’m sure they are going to correct it. It looks like there is an instagram profile listed under “about me” on the website as well. I might reach out and mention this so if they aren’t aware, they can address it :)
Hi, super interesting video. I have a doubt not about deepSNR but about the linear fit that is done at the beginning. If I understand correctly you use the channel with higher mean value as reference. Poking around on the forums I thought I understood that it would be a good idea to use the channel with the lowest SNR as the reference so as to compress the others and not exaggerate the noise of the less represented channel. What advantages are there in your method over the one I mentioned? No criticism, just curious what the difference is! Thank you very much and congratulations again!
Thank you and very good question! All linear fit is trying to do is level the brightness of the images with each other. I have a couple of tutorials where I go through my workflow and how I explain this step is when you’re using a OSC camera you’ll have a tint to it (blue, green or red). It’s hard to see what you’re doing due to this tint and as you can get rid of it through background extraction and a couple of other processes, I separate my channels and use linear fit because this is a beginning stages of color balancing, but the main goal is it gets rid of the tint without affecting the data much. This will then give you an idea of the data’s potential (which is how I word it in my tutorials). As far as highest vs lowest, I have tried it both ways and haven’t noticed a difference in the amount of noise present in the image afterward, although it would make for an interesting video to deep dive into this :) Yes, using the brighter image (higher mean ADU value will brighten the other images that this is applied to, but again, in my workflow I will take care of this within GHS as a first step within GHS. I prefer using brighter with linear fit as it helps me to see what dimmer details may be present that I would like to try to bring out if the data allows. This is preference though for my case. I hope this makes sense, it great question!!
@@Hidden.Light.Photography I thank you so much for your accurate response. I find it super interesting to be able to exchange views on these tiniest details. I think these subtleties in the end are what makes the difference between user who applies processes randomly and one who really knows what he is doing. A comparison on a good dataset would actually be interesting! Ignorantly I had always used an STF with unlinked channels to “hide” the dominant hue when viewing the raw data and delegated the actual removal to the SPCC. But logically using an initial linear fit could also be useful to assist the other processes before SPCC. Thank you again!
Thanks for the shout out Tony and great video as always.
Great new tool. It didn’t work on my RGB images captured with an OSC camera.
Should they be captured via a monochrome camera or do I need to split the channels for it to work?
@@AstroIsland Can you be more specific? It didn't work as in you can't install it? It won't open? Or it does nothing to your image?
@@yervantparnagian5999 sure. The image remains noisy after running the script on my OSC images
You are very welcome and thank you! Great find!!
Are you running on linear data or none linear?
I am amazed at the quality of some of the free options/processes available now. But I wouldn't expect "default" values to generally give you a definitive evaluation of "better" and "worse". I assume the sliders are there for a reason. For example, if you moved the detail slider on NXT, would you get a "better" result?
I agree with you there. The sliders do work to an extent in NoiseXterminator but create other issues. I’m in process of a video going over that as well as go more in depth with deepsnr.
Great vid. Can't wait to try Deep snr and compare to NoiseX
I also need to look into balancing color chanels. I have never done so. Is it used for osc cams too?
Thank you! Yes, color balancing channels is for OSC as well. I also just realized I didn’t put a link to my workflow video 🤣, thank you for the question because I just realized that! Check out this video and you’ll see exactly how to do it :)
th-cam.com/video/LS3Gpx75hhU/w-d-xo.htmlsi=em3u6yPe4221kUIm
Was your image captured using an OSC camera? I saw on the website it only works with RGB images using a monochrome camera?
Great question! Yes, I’m using a OSC camera for this example. It’s a little bit out of bounds of the software’s intention and also accounts for the quirk of the behavior causing it to need to be run early on. I’m in progress of a follow up video going more in depth with it. I did see that it’s not working for you. Can you describe the workflow that you’re using from start to the point of running it?
Looks very promising. I hope you can do OSC also in the future as splitting rgb channels is not in my workflow. I just tried DeepSNR on a OSC file and it doesn’t work.
I am very excited to see where this process goes! Do me a favor. Try running DeepSNR at the very start of your workflow as the first item and let me know if this allows it to work. The image in working on in this video is a OSC
@@Hidden.Light.Photography I ran Deepsnr right after stacking but still no good result.
Just to confirm, you copied both files from the DeepSNR download and pasted in the PixInsight BIN folder? Also, how do you do your stacking and what file type is it?
@@Hidden.Light.Photography Yup, it works but don’t like the result. I stack with Siril and use 32bit fits files. I don’t like what it does to the background. It takes noise and makes little stars of it.
That’s not good. It might be the way Siril is stacking and preprocessing the image. Unfortunately this may be one that needs to be fully developed before using. There are some quirks using my OSC data as well which is done through WBPP in PixInsight that cause me to have to run DeepSNR in the very beginning for best results. I do truly believe it’s worth the wait for this process seeing the absolute potential already.
It works with the Mac but is really really slow. Hopefully they will take advantage of the Mac’s graphics libraries.
Just to make sure, you downloaded the correct PixInsight version? If so, the software is still being developed, so I’m sure they are going to correct it. It looks like there is an instagram profile listed under “about me” on the website as well. I might reach out and mention this so if they aren’t aware, they can address it :)
Hi, super interesting video.
I have a doubt not about deepSNR but about the linear fit that is done at the beginning.
If I understand correctly you use the channel with higher mean value as reference.
Poking around on the forums I thought I understood that it would be a good idea to use the channel with the lowest SNR as the reference so as to compress the others and not exaggerate the noise of the less represented channel.
What advantages are there in your method over the one I mentioned?
No criticism, just curious what the difference is!
Thank you very much and congratulations again!
Thank you and very good question! All linear fit is trying to do is level the brightness of the images with each other. I have a couple of tutorials where I go through my workflow and how I explain this step is when you’re using a OSC camera you’ll have a tint to it (blue, green or red). It’s hard to see what you’re doing due to this tint and as you can get rid of it through background extraction and a couple of other processes, I separate my channels and use linear fit because this is a beginning stages of color balancing, but the main goal is it gets rid of the tint without affecting the data much. This will then give you an idea of the data’s potential (which is how I word it in my tutorials). As far as highest vs lowest, I have tried it both ways and haven’t noticed a difference in the amount of noise present in the image afterward, although it would make for an interesting video to deep dive into this :) Yes, using the brighter image (higher mean ADU value will brighten the other images that this is applied to, but again, in my workflow I will take care of this within GHS as a first step within GHS. I prefer using brighter with linear fit as it helps me to see what dimmer details may be present that I would like to try to bring out if the data allows. This is preference though for my case. I hope this makes sense, it great question!!
@@Hidden.Light.Photography I thank you so much for your accurate response. I find it super interesting to be able to exchange views on these tiniest details.
I think these subtleties in the end are what makes the difference between user who applies processes randomly and one who really knows what he is doing.
A comparison on a good dataset would actually be interesting!
Ignorantly I had always used an STF with unlinked channels to “hide” the dominant hue when viewing the raw data and delegated the actual removal to the SPCC. But logically using an initial linear fit could also be useful to assist the other processes before SPCC.
Thank you again!
I appreciate that! I will start the preparations for a video comparing the two. Would you mind if I credited you for the idea?
@@Hidden.Light.Photography That's great. Thank you very much again! Yes yes, no problem for sure!
Perfect! I will let you know when it’s created and ready to launch :)
Unfortunately, I don't use Pixinsight.
I have some videos in progress for the AVX :) The AVX is actually set up in my studio as we speak