Night Photography Noise Reduction Issues
I've had great success with C1 12 sharpening and reducing noise with my night photography images. These are by their nature quite noisy before processing. No issues with C1 12. But I'm not able to achieve similar results with C1 20.
If I upgrade a C1 12 image to 20 the results are best described as Melted Watercolour with Digital Artefacts. This is most apparent in the foreground. Completely unusable.
If I start from scratch in C1 20 I can get results better than via upgrade - but despite hours of trying different combinations of settings nothing as good as with C1 12. Barely usable.
Perhaps the C1 20 NR engine is attempting to be clever with these long exposure (13 seconds) high ISO (3200) images?
If I upgrade a C1 12 image to 20 the results are best described as Melted Watercolour with Digital Artefacts. This is most apparent in the foreground. Completely unusable.
If I start from scratch in C1 20 I can get results better than via upgrade - but despite hours of trying different combinations of settings nothing as good as with C1 12. Barely usable.
Perhaps the C1 20 NR engine is attempting to be clever with these long exposure (13 seconds) high ISO (3200) images?
0
-
I see much better results on regular light shots with dark areas and high ISO but I think the approach required is a little more complex than with V12 in order to achieve those better results.
I'm not sure I have come close to the best combination of settings so far - especially when sharpening and clarity is involved as well.
I have experimented with one downloaded sample "night shot" but it's probably nothing like you are shooting.
The results were very good but again needed some experimentation and may well be short of their best so far.
I think the source of the file may also be very significant - i.e. the camera/sensor combination..
I would suggest that you share you experiences with the Support Team if you have not already done so.
If you are prepared to share one or two example images I'm sure there are a few people here who would happily take some time to look at them and experiment to see if they get the same impressions or maybe find some alternative processing that works better than you have found so far? Also, perhaps, compare with similar shots from a different camera source but processed in the same way.
HTH.
Grant0 -
mainlyfine wrote:
I've had great success with C1 12 sharpening and reducing noise with my night photography images. These are by their nature quite noisy before processing. No issues with C1 12. But I'm not able to achieve similar results with C1 20.
If I upgrade a C1 12 image to 20 the results are best described as Melted Watercolour with Digital Artefacts. This is most apparent in the foreground. Completely unusable.
If I start from scratch in C1 20 I can get results better than via upgrade - but despite hours of trying different combinations of settings nothing as good as with C1 12. Barely usable.
Perhaps the C1 20 NR engine is attempting to be clever with these long exposure (13 seconds) high ISO (3200) images?
I assume that in CO20, you each time you use the CO20 Engine. I also assume you are aware to upgrade the engine to 20 on older images that you open in CO20. The new noise reduction in Capture One 20 is linked to the new engine.0 -
mainlyfine wrote:
Perhaps the C1 20 NR engine is attempting to be clever with these long exposure (13 seconds) high ISO (3200) images?
Hard to comment without image examples, camera information etc. - but the length of exposure isn't recognised by Capture One, and 3200 ISO isn't high ISO; Capture One 20 provides me with essentially noise-free images way above 3200 ISO, albeit not shooting at night.
My point being - more information, please.0 -
Thanks all.
The example images are taking with a Fujifilm X-T2 using a Rokinon 21mm f1.4 lens.
A stack of RAW images is imported in to C1, lens corrections made and TIFFs exported to Starry Landscape Stacker.
Usually many resulting stacks are stitched together to create a pano. In this example I have a single stack.
The resulting TIFF is then processed in C1.
Initial TIFF (large)
https://www.dropbox.com/s/9qd7pd2a7ccx3tp/2019.08.03%20-%20_XT28674%20-%20100%20Full-Min%20Horizon%20Noise.tif?dl=0
Edited using v12 Engine
https://www.dropbox.com/s/6ulcv7es67vkp0s/Night%20Photography%20-%20C1%20v12.jpg?dl=0
Upgrade from v12 Engine to v20 Engine (no further editing). Foreground is Melted Watercolour
https://www.dropbox.com/s/mfpus0c62eqz5br/Night%20Photography%20-%20C1%20v20%20Upgraded.jpg?dl=0
With v12 my settings were:
Background (entire image): Luminance 50, Detail 50, Sharpening Amount 100, Radius 0.5, Threshold 1
Foreground (masked): Luminance 30, Detail 25, Sharpening Amount 100, Radius 0.5, Threshold 1
I've continued to work with v20 and have gotten close to v12 - but still have a bit more noise and a bit less sharpness. Most noticeable when printed (the entire image - not this snippet).
Any additional NR resulted in more detail lost. More sharpening resulted in artefacts.
Best v20:
Background: Luminance 30, Detail 75, Sharpening Amount 50, Radius 0.2, Threshold 1
Foreground: Luminance 0, Detail 50, Sharpening Amount 0, Radius 0.5, Threshold 10 -
mainlyfine wrote:
Thanks all.
The example images are taking with a Fujifilm X-T2 using a Rokinon 21mm f1.4 lens.
A stack of RAW images is imported in to C1, lens corrections made and TIFFs exported to Starry Landscape Stacker.
Usually many resulting stacks are stitched together to create a pano. In this example I have a single stack.
The resulting TIFF is then processed in C1.
Initial TIFF (large)
https://www.dropbox.com/s/9qd7pd2a7ccx3tp/2019.08.03%20-%20_XT28674%20-%20100%20Full-Min%20Horizon%20Noise.tif?dl=0
Edited using v12 Engine
https://www.dropbox.com/s/6ulcv7es67vkp0s/Night%20Photography%20-%20C1%20v12.jpg?dl=0
Upgrade from v12 Engine to v20 Engine (no further editing). Foreground is Melted Watercolour
https://www.dropbox.com/s/mfpus0c62eqz5br/Night%20Photography%20-%20C1%20v20%20Upgraded.jpg?dl=0
With v12 my settings were:
Background (entire image): Luminance 50, Detail 50, Sharpening Amount 100, Radius 0.5, Threshold 1
Foreground (masked): Luminance 30, Detail 25, Sharpening Amount 100, Radius 0.5, Threshold 1
I've continued to work with v20 and have gotten close to v12 - but still have a bit more noise and a bit less sharpness. Most noticeable when printed (the entire image - not this snippet).
Any additional NR resulted in more detail lost. More sharpening resulted in artefacts.
Best v20:
Background: Luminance 30, Detail 75, Sharpening Amount 50, Radius 0.2, Threshold 1
Foreground: Luminance 0, Detail 50, Sharpening Amount 0, Radius 0.5, Threshold 1
OK, some initial thoughts come to mind.
If you leave the processing engine as V12 when using V20 do you see the same results as V12 or not? (Is that what you did with the sample images? I was not sure if the V12 was pure V12 or V20 using the V12 engine.)
The Tiff will start for some very generic settings whereas working with an original image will likely have a different starting point that takes into account some assessment of the information in the file - especially a RAW file - related to the camera, sensor, ISO settings and so on. Exactly what is used is unclear BUT the default slider settings are not absolute values that are the same for all images. They are, where possible, a 'best estimate' of what is likely to work well for that particular image.
My guess is that for a stacked TIFF with a lot of combined noise ... trying to assess it is not going to be easy.
Your Rokinon lens is, perhaps, a Samyang derivative?
If so I have a 14mm Samyang and it's a great lens but with a little special distortion that normal correction cannot correct. I have no idea if a21mm would have the same sort of challenge but it seems likely and presumably the lens does not provide any correction data?
If so I have to wonder about the accuracy possible from the stacking activity and how that might affect the noise result.
I really don't know if it would matter but something suggest to me that, when dealing with fine details, it might. What is your observation about that?
And finally, for now, based on what little I have experimented with rather dark night time images (as mentioned previously) I think you may need to experiment with some fairly extreme adjustments, compared to usual expectations, for this type of image. Just a hunch at the moment - I'll take a look at your sample tomorrow.
You probably also need to check the results at 100% viewing size when assessing the corrections required and then consider what might be required for smaller sizes later.
There is, of course, no certainty that the intent behind C1's NR approach will be suitable for this sort of image construction - but let's see where we can get to.
Grant0 -
Thanks Grant
The v12 image is "pure v12". All sharpening and NR changes were made in v12 before the v20 was installed.
One of the original RAF files is here:
https://www.dropbox.com/s/czpj03bj4mjsizh/2019.08.03%20-%20_XT28674.RAF?dl=0
Rokinon/Samyang are the same. Based on experimentation I created a Lens Correction Preset that seems to perform well.
Chromatic Aberration, Diffraction Correction, Hide Distorted Areas all Ticked
Distortion 7, Sharpness Falloff 20, Light Falloff 35
Capture One performs the initial RAW file processing with the benefit of all the RAF file data before exporting for stacking.
Starry Landscape Stacker creates a mask and treats the sky and foreground portions of images differently (allowing for movement in the sky). My experience is significantly reduced noise (based on the number of images in the stack) with limited if any reduction in details.
Agreed on the need to confirm at 100%. The v12 and v20 previews look virtually identical.
I've used a similar workflow with Lightroom, Luminar and Capture One and v12 was the best. I'm hoping I can get back there with v20.
Thanks again for your ideas and insights.0 -
I have had a quick look at your original RAF file.
The Exposure Histogram seems to be suggestiing that it is 3 or maybe 4 stops underexposed overall. That would of course be related to foreground and midground and somewhat expected for the type of shot.
There are a few thoughts that come with that in terms of process the RAW file. I#m still working through them but one thing to ask is whether you set the Base Characteristics to Linear Response Curve.
This gives a flat result but it's something I use all of the time and I think it gives a better degree of control for the sort of adjustments required here.
If you are already using it then great. If not try it and see what you think.
Grant0
La publicación no admite más comentarios.
Comentarios
7 comentarios