The GetDPI Photography Forum

Great to see you here. Join our insightful photographic forum today and start tapping into a huge wealth of photographic knowledge. Completing our simple registration process will allow you to gain access to exclusive content, add your own topics and posts, share your work and connect with other members through your own private inbox! And don’t forget to say hi!

Quick example of Frame Averaging on IQ4

Paul2660

Well-known member
Frame Averaging side by side .jpgFrame Averaging side by side close up.jpg



This is from a recent shoot I did in Clarendon Arkansas. I love old truss bridges and there is a real wonder built in 1870 on the Union Pacific line there.

I have shot this bridge many times but never with a sunset, and wanted to try it. The conditions were perfect for Frame Averaging, as there was no wind. I had a very bright sun setting on the left side and almost no light to the right. I choose to expose for the highlights and try to pull up the shadows later. Normally this would be 4 shot bracket, but the ability frame averaging allows is really amazing. I have attached two screens shots. First shows the final image worked in C1 and the image without any adjustments, and the 2nd is just a crop, showing that the bridge, and the area below it was basically black. I have never found any camera that can pull up from black and still allow highlight recovery, no Phase Back for sure as the area under the bridge would be full of noise.

You can easily see plenty of details from the area under the bridge, which was in low or if any ambient light at all.

I like the grudge look on structures such as this, so I apologize in advance to those who prefer a different look. My point is really just to show how much recovery (clean recovery) is possible with Frame Averaging.

I have been hard on P1 as many other due to the missing firmware tools that the IQ4 has not received that previous P1 backs had and no Capture Pilot solution. But to be honest I can forgive a lot with this type of shadow recovery. I have never seen anything like this before from a single frame.

Shot with Area rm3di and 32mm Rodenstock, no CF, F11 (diffraction be damned) Note, slight haloing on the fine guard wire on the bridge rail, missed this at first easily fixed with "halo" suppression in C1

Paul C
 

alatreille

Member
Nice work Paul

It's a great tool in the right circumstances.

How many images and for how long did you 'average' this over?

Cheers

Andrew
 

Paul2660

Well-known member
Looking at the exif info in C1.

Shutter speed was 1/13 of a second, ISO 50. Frame averaging info is 2.36 second.

So if I understand this, I shot enough frames at 1/13 of a second to create a 2.36 exposure.

I hate to admit it, but when using this tool I tend to throw settings at the wall till I find something that sticks. I ended up later that night using FA for a 2 minute exposure, which I will add to the MF images later today.

I choose to stay at base ISO of 50 based on more my experience of taking longer exposures always seems to work best at base ISO. I realize that with FA, many others have shown that they can get great results pushing ISO. However my biggest concern was the issue of blowing the highlight of the setting sun, and I felt a higher ISO would tend to blow out the left side. So I went the other way, exposed for the sky and when hoped to push the shadows in post. The results were more than I expected.

Paul C
 

Craig Stocks

Well-known member
Great example.

I agree that the IQ4150 really is a stellar product, it’s a shame it’s crippled by missing features and Phase One’s apparent incompetence with firmware updates. Frame averaging, live view focus mask and Ethernet tethering have proven to be very useful for my workflow but I really miss a Capture Pilot solution.
 

dougpeterson

Workshop Member
Looking at the exif info in C1.

Shutter speed was 1/13 of a second, ISO 50. Frame averaging info is 2.36 second.

So if I understand this, I shot enough frames at 1/13 of a second to create a 2.36 exposure.

I hate to admit it, but when using this tool I tend to throw settings at the wall till I find something that sticks. I ended up later that night using FA for a 2 minute exposure, which I will add to the MF images later today.

I choose to stay at base ISO of 50 based on more my experience of taking longer exposures always seems to work best at base ISO. I realize that with FA, many others have shown that they can get great results pushing ISO. However my biggest concern was the issue of blowing the highlight of the setting sun, and I felt a higher ISO would tend to blow out the left side. So I went the other way, exposed for the sky and when hoped to push the shadows in post. The results were more than I expected.

Paul C
You can find the details of the frame averaging in the metadata tab in Description. The more frames averaged the more impressive the suppression of noise in the very very deep shadows.

For those playing catchup (i.e. that are not previously familiar with the feature) you can learn more about frame averaging here:
https://phaseoneiq4.com/frame-averaging-for-infinite-dynamic-range/
 

Paul2660

Well-known member
Frame Av side by side B.jpgFrame Av side by side B close up .jpgFrame Av side by side B close up 2.jpg

One more example, from same evening, about 45 minutes later.

These are from a single image 2 minutes long, Frame Averaged from a 2 minute series of 15" second exposures. ISO 50. The detail in the crane and trees on far bank is most impressive. Almost noise free.

Paul C
 

Steve Hendrix

Well-known member
Paul, this is a really good example of needing to get the exposure down to effectively capture the highlight, but then needing to pull shadows up dramatically as a result.

Below I am re-linking Brad Kaye's examination of Frame Averaging techniques and some of the myriad ways it can be beneficial.

IQ4 Frame Averaging Is Finally Here


Steve Hendrix/CI
 

Boinger

Active member
Yep really good posts.

Another tip I read on another post somewhere but I thought I would mention here too.

If you create a custom style of say Shadows boosted 100 or 50 or whatever your flavor. And apply to capture.

Then you can get a rough idea of what the image will look like on the iq4 150 lcd screen.
 

f8orbust

Active member
Interesting thread, but I don't quite understand why so many images need to be captured. Is this determined by the back, or can the user set the number of 'steps' ?

I only ask because Guillermo Luijk showed - 10 years ago now - that there's nothing to be gained by shooting more than 4 images - and in fact he determined that 3 images 0EV, +2EV, and +4EV was perfect in most situations just so long as the base exposure (the '0') didn't clip any highlights.

Obviously the fewer images you need to capture the less chance there is for movement between images degrading the final (composite) image quality.

Jim
 

steve_cor

Member
Jim,
There are no number of 'steps' because it's not HDR. All the shots are at the same exposure.

There would have been noise in the shadows. The idea behind the frame averaging is the noise will be removed since it is in different places in every shot.



--Steve.
 

mristuccia

Well-known member
Interesting thread, but I don't quite understand why so many images need to be captured. Is this determined by the back, or can the user set the number of 'steps' ?

I only ask because Guillermo Luijk showed - 10 years ago now - that there's nothing to be gained by shooting more than 4 images - and in fact he determined that 3 images 0EV, +2EV, and +4EV was perfect in most situations just so long as the base exposure (the '0') didn't clip any highlights.

Obviously the fewer images you need to capture the less chance there is for movement between images degrading the final (composite) image quality.

Jim
Jim,

I think this is another technique to achieve "almost" the same result.
You're referring to the usual HDR technique, which extends the dynamic range by putting different zones of the subject in the center of the histogram in different shots, and then by taking this best (central) part out every shot and putting everything together to form the final image.

In the case of the Frame Averaging we take exactly the same shot with the same exposure many times. By averaging each pixels (in the same position) of each shot we rule out the random noise (aka Shot Noise or Photon Noise). So in the end the image is so clean that we can push the shadows without revealing any noise. In this case we rely on the native dynamic range of the digital back. Only... we make a little bit of clean-up. :)

Through the first technique we avoid noise by putting the area on the center of the histogram, where the Signal to Noise ratio is at his best thanks to the native sensor's technology.
By means of the second technique we improve the Signal to Noise ratio everywhere, and mostly on the shadows where it is lower, but we leave them (the shadows) where they usually are, which is in the leftmost area of the histogram.
 
Last edited:

f8orbust

Active member
Thanks Marco. So the way P1 implements it is basically akin to doing the following in PS:

Import images as a stack in PS
Auto align the layers
Convert all the layers to a smart object
Then select layer > smart object > stack mode > median

So in theory the more images you have, the cleaner the result, but in the real world once you get above 3 images the effect tends to be less noticeable, with 5 images often regarded as the 'sweet spot'.
 

mristuccia

Well-known member
Thanks Marco. So the way P1 implements it is basically akin to doing the following in PS:

Import images as a stack in PS
Auto align the layers
Convert all the layers to a smart object
Then select layer > smart object > stack mode > median

So in theory the more images you have, the cleaner the result, but in the real world once you get above 3 images the effect tends to be less noticeable, with 5 images often regarded as the 'sweet spot'.
I don't know exactly how P1 implements this, but you're right. The manual technique is the one you described.

Regarding the best number of shots, I think it also depends on your ISO setting. Theoretically doubling the number of shots it's like having one ISO stop less in terms of noise.
So, for example, if you average 16 shots taken at ISO 1600 you should get a noise figure which is more or less the one you would have gotten with one shot at ISO 100.
 

Christopher

Active member
The theory regarding number of shots isn’t correct. While it’s true 5 are ok, 20 are so much better it’s incredible. I’m still testing it more and more but it’s incredible how completely noise free a image with 100 images is.
 

Paul2660

Well-known member
As a user of stacking for years in Astro work, I am very familiar with Guillermo Luijk's work and others. Many forget that the all the stack modes in Photoshop were not created by Adobe, but by a end user (I forget his name), and Adobe found them impressive enough to add formally to Photoshop, several years ago.

I rarely stacked for just noise reduction, however in night work, the stacking always creates a much cleaner sky, either in star trail work or Milky Way.

However where the IQ4 totally impresses me, is how it is processing out the raw files into a single FA file in the time it takes for the total exposure. My 2nd example was a total of a 2 minute exposure, with 15 second stacks. At the end of 2 minutes, you are done, there is no additional time taken to write the file, process the file, etc. You are ready to take the next exposure.

Doing this manually on a Mac or PC, is very time consuming.

1. Convert each file from raw to tif
2. Auto align all layers
3. Convert to smart Object
4. Work the file.

Attempting to convert 20 IQ4 tifs to a Smart Object alone would be something I would not want to attempt as Photoshop will start to choke on 50 Nikon D810 images (star trail work). The Alignment steps can work or not work creating yet another time consuming step to get all he parts of the single image combined. Yet this is on a PC/Mac, with 64 GB of ram, and a i7 8 core or 6 core Xeon Mac.

The IQ4, still lacking in many of the features that Phase made "standard" with previous CMOS backs, does show that there is something different under the hood and it's quite capable of handling large processing requirements. Sure it's not converting the files to tifs, and 16 bit output as the PC stacking method is but it's still working on a pretty high number of images and able to complete the process as within the set exposure time.

As for the more or less images, I can state that in some cases in normal bright light, only averaging 2 images can make a big difference in areas that were dark thus contained noise. I don't have any method when I use the tool i.e set number of images taken for a certain amount of exposure. Since the process doesn't take any additional time from the normal exposure time selected, you can take a lot of images as the final image is only being stored as a normal raw file on the card.

The issues around ES and motion etc. to me are moot. If you have motion you will have issues with any method of frame averaging, be the Phase One solution or manually in Photoshop as Photoshop IMO can't align the smaller details like leaves and branches any better.

For any long exposure single exposure motion will be an issue also, but it's clear to me that a 1 second single shot vs a 1 second frame averaged shot the single non frame averaged image will contain more noise. Since the process is so easy to implement on the back, I just make it a point to always shoot a frame averaged image as there will always be areas that will not have motion and will gain from the FA process.

Paul C
 

Christopher

Active member
Super dirty example of single shot (top left), 2 stack (top right), 5 stack (bottom left) and 20 stack (bottom right) - Second picture shows exposure before lifting the shadows.

This is not a normal exposure, just showing what would happen if you have something very bright and need to expose for that. It shows how much information is in the shadows.

Stacking_Detail.JPG

Stacking_Detail_Before.JPG
 

MGrayson

Subscriber and Workshop Member
Software to average RAW files is a few lines of code. Just because PS has to do a ton of processing before averaging doesn't mean it's miraculous that Phase does it in the back. Having said that, those few lines of code aren't lying around.

The reason it's much easier than it looks is that the running Mean can be computed without storing more than one frame.
For k = 1 to number of frames
Mean = ((k-1) Mean + New Frame)/k
All done before RAW conversion. No alignment, no processing. It's easy. Median would be harder, as there is no incremental update to Median.

True, you'd need to store all those frames on the card, so it saves both card space AND write time. Don't get me wrong, it's a Good Thing, and every modern camera should have it.

Kudos to Phase One.

Matt
 
Top