p.32 #1 · Sigma 14-24mm f/2.8 DG DN first impressions
Minatureman13 wrote:
Been following this thread, and I'm kind of excited about this mean stacking that your talking about. Those images look fantastic.
Just to clarify on how you did this... For the first image, is this what you did?
1. Take 70 photos with the background in focus and exposed as well as possible
2. Take 16 photos with the foreground in focus and exposed as well as possible
3. mean stack the 70 background photos to get a single clean background image
4. post process that single resulting image, focusing on the background
5. mean stack the 16 foreground images to get a single clean foreground image
6. post process the single resulting image, focusing on the background
7. then combine the resulting two clean images
However, for the improved SNR (Signal to Noise Ratio) alone, 16 images is the most practical number IMO. So, even if you are using ND filters, shooting 16 images and "mean" averaging them, will improve your images dramatically.
Basically the most drastic improvement happens when averaging 4 images since it would cut random noise in half, so it's the same as shooting at ISO 50. (Real ISO 50, not the fake one offered by the camera, which is basically just adding 1EV to ISO 100 and underexposing to get to ISO 50)
Averaging images improves the image "signal" and the reduction in noise can be calculated using the below formula:
"Noise" drops by the square root of the number of images averaged
The SNR improvement diminishes further as we continue to double the number of averaged images though. Using 16 images already gives us 25% of the original random noise. So, if you are shooting at ISO 100, it will be the same as shooting at ISO 25. You can go further from there but we start getting diminishing returns.
Now, for mimicking the ND effect in our images, the more images averaged the stronger the ND effect.
For the examples on this page, I didn't use any ND filter so 256 images mimic a 8-stop ND filter and 70 images a 6-stop filter.
Here is a quick table:
Starting with no filter:
1 stop (2 shots averaged)
2 stops (4 shots averaged)
3 stops (8 shots averaged)
4 stops (16 shots averaged)
5 stops (32 shots averaged)
6 stops (64 shots averaged)
7 stops (128 shots averaged)
8 stops (256 shots averaged)
Starting with a 4-stop filter:
5 stops (2 shots averaged)
6 stops (4 shots averaged)
7 stops (8 shots averaged)
8 stops (16 shots averaged)
9 stops (32 shots averaged)
10 stop (64 shots averaged)
Starting with a 6-stop filter:
7 stops (2 shots averaged)
8 stops (4 shots averaged)
9 stops (8 shots averaged)
10 stops (16 shots averaged)
Important: If you are Mean Averaging more than 16 images, DO NOT bring them from Lightroom as layers. This will slow down the process considerably (or even crash your computer)
The best way is to load the images into a stack using Photoshop's File/Scripts/Load Files into Stack.
Browse your files and check on: Create Smart Object after loading layers
Once this is complete, go to: Layer/Smart Objects/Stack Mode/Mean
Using the method above, my system can averaged 256 images without a hiccup. (32 images only takes 1min 30s from start to end)
This is a bit off topic but this info could be useful as an alternative to using ND filters with the Sigma 14-24 DG DN or any other lens when aiming for the highest IQ.
p.32 #2 · Sigma 14-24mm f/2.8 DG DN first impressions
Fred Miranda wrote:
That's exactly the process.
However, for the improved SNR (Signal to Noise Ratio) alone, 16 images is the most practical number IMO. So, even if you are using ND filters, shooting 16 images and "mean" averaging them, will improve your images dramatically.
Basically the most drastic improvement happens when averaging 4 images since it would cut random noise in half, so it's the same as shooting at ISO 50. (Real ISO 50, not the fake one offered by the camera, which is basically just adding 1EV to ISO 100 and underexposing to get to ISO 50)
Averaging images improves the image "signal" and the reduction in noise can be calculated using the below formula:
"Noise" drops by the square root of the number of images averaged
The SNR improvement diminishes further as we continue to double the number of averaged images though. Using 16 images already gives us 25% of the original random noise. So, if you are shooting at ISO 100, it will be the same as shooting at ISO 25. You can go further from there but we start getting diminishing returns.
Now, for mimicking the ND effect in our images, the more images averaged the stronger the ND effect.
For the examples on this page, I didn't use any ND filter so 256 images mimic a 8-stop ND filter and 70 images a 6-stop filter.
Here is a quick table:
Starting with no filter:
1 stop (2 shots averaged)
2 stops (4 shots averaged)
3 stops (8 shots averaged)
4 stops (16 shots averaged)
5 stops (32 shots averaged)
6 stops (64 shots averaged)
7 stops (128 shots averaged)
8 stops (256 shots averaged)
Starting with a 4-stop filter:
5 stops (2 shots averaged)
6 stops (4 shots averaged)
7 stops (8 shots averaged)
8 stops (16 shots averaged)
9 stops (32 shots averaged)
10 stop (64 shots averaged)
Starting with a 6-stop filter:
7 stops (2 shots averaged)
8 stops (4 shots averaged)
9 stops (8 shots averaged)
10 stops (16 shots averaged)
Important: If you are Mean Averaging more than 16 images, DO NOT bring them from Lightroom as layers. This will slow down the process considerably (or even crash your computer)
The best way is to load the images into a stack using Photoshop's File/Scripts/Load Files into Stack.
Browse your files and check on: Create Smart Object after loading layers
Once this is complete, go to: Layer/Smart Objects/Stack Mode/Mean
Using the above way, my system can averaged 256 images without a hiccup. (32 images only take 1min 30s from start to end)
This is a bit off topic but this info could be useful as an alternative to using ND filters with the Sigma 14-24 DG DN or any other lens when aiming for the highest IQ....Show more →
Huh. Interesting tip about not using LR to send them to PS as layers, which is what I usually do.
Is there a number of layers where this starts to become really important in your experience, Fred?
p.32 #3 · Sigma 14-24mm f/2.8 DG DN first impressions
Fred Miranda wrote:
That's exactly the process.
However, for the improved SNR (Signal to Noise Ratio) alone, 16 images is the most practical number IMO. So, even if you are using ND filters, shooting 16 images and "mean" averaging them, will improve your images dramatically.
Basically the most drastic improvement happens when averaging 4 images since it would cut random noise in half, so it's the same as shooting at ISO 50. (Real ISO 50, not the fake one offered by the camera, which is basically just adding 1EV to ISO 100 and underexposing to get to ISO 50)
Averaging images improves the image "signal" and the reduction in noise can be calculated using the below formula:
"Noise" drops by the square root of the number of images averaged
The SNR improvement diminishes further as we continue to double the number of averaged images though. Using 16 images already gives us 25% of the original random noise. So, if you are shooting at ISO 100, it will be the same as shooting at ISO 25. You can go further from there but we start getting diminishing returns.
Now, for mimicking the ND effect in our images, the more images averaged the stronger the ND effect.
For the examples on this page, I didn't use any ND filter so 256 images mimic a 8-stop ND filter and 70 images a 6-stop filter.
Here is a quick table:
Starting with no filter:
1 stop (2 shots averaged)
2 stops (4 shots averaged)
3 stops (8 shots averaged)
4 stops (16 shots averaged)
5 stops (32 shots averaged)
6 stops (64 shots averaged)
7 stops (128 shots averaged)
8 stops (256 shots averaged)
Starting with a 4-stop filter:
5 stops (2 shots averaged)
6 stops (4 shots averaged)
7 stops (8 shots averaged)
8 stops (16 shots averaged)
9 stops (32 shots averaged)
10 stop (64 shots averaged)
Starting with a 6-stop filter:
7 stops (2 shots averaged)
8 stops (4 shots averaged)
9 stops (8 shots averaged)
10 stops (16 shots averaged)
Important: If you are Mean Averaging more than 16 images, DO NOT bring them from Lightroom as layers. This will slow down the process considerably (or even crash your computer)
The best way is to load the images into a stack using Photoshop's File/Scripts/Load Files into Stack.
Browse your files and check on: Create Smart Object after loading layers
Once this is complete, go to: Layer/Smart Objects/Stack Mode/Mean
Using the method above, my system can averaged 256 images without a hiccup. (32 images only takes 1min 30s from start to end)
This is a bit off topic but this info could be useful as an alternative to using ND filters with the Sigma 14-24 DG DN or any other lens when aiming for the highest IQ....Show more →
It will no doubt depend on how much you are needing to push your shadows, but I find that even using just 8 shots gives great practical SNR improvements.
On a related note, I was previously under the impression that median averaging would give more SNR benefit than using the mean. But according to this that's only true in terms of getting rid of outlier noise. For random noise, the median is actually about 20% less effective than the mean, so the latter is the better stacking mode for this purpose.
p.32 #5 · Sigma 14-24mm f/2.8 DG DN first impressions
rvh23 wrote:
It will no doubt depend on how much you are needing to push your shadows, but I find that even using just 8 shots gives great practical SNR improvements.
On a related note, I was previously under the impression that median averaging would give more SNR benefit than using the mean. But according to this that's only true in terms of getting rid of outlier noise. For random noise, the median is actually about 20% less effective than the mean, so the latter is the better stacking mode for this purpose.
I've compared 16 averaged images using Mean and Median and noticed that the latter is not as clean (I think 20% sounds right) and not as smooth as the former. For mimicking the ND effect, mean averaging is the best choice.
However, if you want to remove moving subjects from the frame, median is better than mean. Travel photography would be good candidate for using median averaging in order to remove people walking around monuments, street, etc.
p.32 #7 · Sigma 14-24mm f/2.8 DG DN first impressions
DavidBM wrote:
Processing speed! AFAIK it should make no difference to NR whether you send to layers from LR or use load files from within PS...
I don't have the fastest machine out there. It's a Mac Mini 3.2 GHz 6-core i7 with 64 GB DDR4 ram with a Vega 64 eGPU. For me, it's doable sending 32 images from LR to PS as layers but it takes a very long time to load, create a smart object and mean average the layers. (about 25min)
Loading the 'unprocessed' RAW files from PS instead, only takes about 2 minutes, so it's a very considerable difference.
I'd say that with more than 10-12 images, it's not worth doing the LR export to PS as layers. You can push it to 16 files if you don't mind the wait. (or perhaps you have a faster system)
I usually load the files as a stack straight from the memory card...
First, in PS, I do:
File/Import/Image from Device
This will show me all the thumbnail images from my memory card which allows me to take note of which file starts the sequence and which ends it.
Then I do: Scrips/Load files into a Stack..and run an action that converts the layers into a smart object and run the mean stack on it. It's a pretty simple workflow.
That's the file I bring to LR to start my editing.
Nice review. It points out two characteristics that make this lens even more desirable. The almost complete lack of axial and lateral CA and the round specular highlights off-axis even wide open.
p.32 #9 · Sigma 14-24mm f/2.8 DG DN first impressions
I got my Lee 1.2 filter sheet from Adorama today, and man, I have to say, it was quite crusty.
Some of the marks seem to be scratches in the sheet that I can't remove with a cleaning cloth.
Is this normal?
p.32 #10 · Sigma 14-24mm f/2.8 DG DN first impressions
SDTrojan wrote:
I got my Lee 1.2 filter sheet from Adorama today, and man, I have to say, it was quite crusty.
Some of the marks seem to be scratches in the sheet that I can't remove with a cleaning cloth.
Is this normal?
p.32 #11 · Sigma 14-24mm f/2.8 DG DN first impressions
SDTrojan wrote:
I got my Lee 1.2 filter sheet from Adorama today, and man, I have to say, it was quite crusty.
Some of the marks seem to be scratches in the sheet that I can't remove with a cleaning cloth.
Is this normal?
Dang, ugly.
I think those filter sheets are intended for modifying light sources. They aren't supposed to be optical quality (and clearly they aren't)
p.32 #14 · Sigma 14-24mm f/2.8 DG DN first impressions
Fred Miranda wrote:
That looks really bad. Where did you get it from?
Adorama. I was wondering if this is normal. No amount of scrubbing with water and a microfiber cloth gets the marks off. They appear to be physical scratches.
p.32 #17 · Sigma 14-24mm f/2.8 DG DN first impressions
MikeEvangelist wrote:
I feel your pain. I've been in that spot a few time. I was actually surprised to get a decent offer so soon from GT.
That's a great price but be aware I've had issues returning defective lenses using Greentoe in the past. (Unless it's from a reputable store in their network)
p.32 #18 · Sigma 14-24mm f/2.8 DG DN first impressions
Fred Miranda wrote:
That's a great price but be aware I've had issues returning defective lenses using Greentoe in the past. (Unless it's from a reputable store in their network)
I NEVER buy lenses from Greentoe, for exactly that reason. I only consider them for items with a very low probability of sub-optimal performance. And even then, unless I am getting a super deal and there is virtually no possibility of my wanting to return the item, I prefer to go with a good retailer.
The difficulty of returns is a very significant weakness for Greentoe purchases.
p.32 #19 · Sigma 14-24mm f/2.8 DG DN first impressions
chiron wrote:
I NEVER buy lenses from Greentoe, for exactly that reason. I only consider them for items with a very low probability of sub-optimal performance. And even then, unless I am getting a super deal and there is virtually no possibility of my wanting to return the item, I prefer to go with a good retailer.
The difficulty of returns is a very significant weakness for Greentoe purchases.
I share those concerns, but have had pretty good luck with GT stuff. Bought 5 lenses that way; got one bad one, and the return process was a little slow, but otherwise fine.
p.32 #20 · Sigma 14-24mm f/2.8 DG DN first impressions
SDTrojan wrote:
I got my Lee 1.2 filter sheet from Adorama today, and man, I have to say, it was quite crusty.
Some of the marks seem to be scratches in the sheet that I can't remove with a cleaning cloth.
Is this normal?
