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!

IQ4150 Frame Averaging Comparisons

Craig Stocks

Well-known member
I've spent a fair amount of time this morning experimenting with frame averaging on the IQ4150. I posted yesterday that frame averaged results are far superior to single frames in terms of noise but I also wanted to compare to traditionally stacked and averaged images using Photoshop.

Observations and questions:

Frame averaging can greatly reduce noise which allows a lot of latitude to lift exposure and shadows without incurring excessive noise.

It does not (my opinion) increase dynamic range. Bright highlights and deep shadows are equally recoverable in a single frame or FA, there's just less noise when averaged. Exposing for the highlights and lifting shadows in post can yield a clean image, but it doesn't automatically do any tone mapping.

Taking individual frames and averaging in Photoshop is no worse and may be slightly better than averaging in the IQ4150. Note that in this test I pushed exposure, etc. in Capture One working with RAW files and then stacked the resulting images.

I have seen some magenta artifacts when bright highlights are present for only a small portion of the total frame average exposures. I'm still exploring this issue and working with my dealer.

The explanation from Phase One is still somewhat convoluted. I'm curious if it does the calculations on RAW date or after demosaicing. Stacking in Photoshop has to be done on a demosaiced image.

Samples and test:

I arranged a simple, high contrast scene in my studio. All samples were captured at ISO 50 @ 1 second and I used 10-frame stacks for comparison. I processed all frames in Capture One by setting the WB to Tungsten, and then pushed Exposure, Highlights and Shadows to the max. I also pushed contrast almost to the max.

I exported the test images as 16-bit ProPhoto PSDs and stacked in Mean stacking mode in Photoshop for comparison.

The samples here were cropped to 100% and saved as sRGB JPEGs.
 

Attachments

Craig Stocks

Well-known member
Not much of a surprise - processing a PSD in Photoshop is MUCH worse than processing the RAW in C1. For this comparison I output the original unadjusted 16-bit ProPhoto files from C1 and then stacked in Photoshop. I then attempted to brighten the images in PS. Single frame, stacked in Photoshop and frame averaged in the camera all look equally bad.

Brighten shadows in RAW and then stack in PS to reduce noise - GOOD

Stack in PS to reduce noise then brighten shadows in PS - BAD
 

Attachments

SrMphoto

Well-known member
Not much of a surprise - processing a PSD in Photoshop is MUCH worse than processing the RAW in C1. For this comparison I output the original unadjusted 16-bit ProPhoto files from C1 and then stacked in Photoshop. I then attempted to brighten the images in PS. Single frame, stacked in Photoshop and frame averaged in the camera all look equally bad.

Brighten shadows in RAW and then stack in PS to reduce noise - GOOD

Stack in PS to reduce noise then brighten shadows in PS - BAD
In my experience, it is always better to brighten shadows in RAW than in TIFF, regardless if you do raw manipulation in C1, LR or ACR.
 

dougpeterson

Workshop Member
The explanation from Phase One is still somewhat convoluted. I'm curious if it does the calculations on RAW date or after demosaicing. Stacking in Photoshop has to be done on a demosaiced image.
I agree Phase One's answer is hard to parse.

Given that the result is a raw file (with a file size commensurate to true raw data, not RGB/demosaic'd data) it's clear the math is done on the raw pixel level, not after demosaicing.
 

SrMphoto

Well-known member
I've spent a fair amount of time this morning experimenting with frame averaging on the IQ4150. I posted yesterday that frame averaged results are far superior to single frames in terms of noise but I also wanted to compare to traditionally stacked and averaged images using Photoshop.
Thanks for the tests. Have you tried varying the number of frames? I wonder how many are necessary to make the noise disappear entirely. My D850 can do multiple exposures and generates an averaged raw file (14-bit, max 10 frames). I am not comparing the image quality of D850 with IQ4150 here. I wonder if I should use my D850's multiple exposure more often to extract the last bit of image quality.
 

dchew

Well-known member
Here is another example, my first experiment. 0.5 sec, f/11, ISO 50. ~ 4 stops underexposed. This is 100% crop after +4 Exposure and 30 Shadows. Std left, 20 frames FA right.

Dave

 

Paul2660

Well-known member
Dave

Do you use the Topaz Denoise AI software.

Just curious is you run the denoise on the none frame averaging shot if the end result will be similar? Topaz has really nailed noise reduction without loss of details.

Paul C
 

dchew

Well-known member
Dave

Do you use the Topaz Denoise AI software.

Just curious is you run the denoise on the none frame averaging shot if the end result will be similar? Topaz has really nailed noise reduction without loss of details.

Paul C
Hi Paul,
No I do not, that is a good point. In fact, it has been several years since I've done any noise reduction at all on files above my defaults in C1/LR. That goes for other camera systems too: IQ180, IQ3100, a7rii and even the 18mp Leica Monochrom. I know that sounds crazy; I just never see noise anymore in a print. Well, maybe it is more accurate to say I haven't worried about it in a print. Perhaps it has been visible but I have not taken the time to notice. I've made several 90" prints, but most of those are two-image stitched panos and they were all at base ISO.

I am not too sure about how I am going to use this feature. On the one hand:
  • It is pretty easy to use, so why not use it whenever possible.
  • When files are really pushed you can, at least on-screen, see a clear difference.
On the other hand, I'm not sure how often this noise benefit is going to "expose itself" in a print. And as you point out, I could at least make a minimal effort to reduce that single-frame noise! There are, of course, other benefits to FA like eliminating the need to carry a bunch of ND filters and getting rid of things like those crazy birds flying around Stokksnes.
:thumbup:

Dave
 

dougpeterson

Workshop Member
An infinite number. How many to make the noise invisible at the desired viewing distance? That's another question.

😉
I would posit that "inoffensive" is a better standard than invisible. That is, of course, subjective, but seems to be the final result that you should care about.

Some images benefit from heavy grain. Some do not. The question is what camera, technique, and post processing will get you the image you want with the amount of grain that fits the image (whether high, low, or near zero).
 

Paul2660

Well-known member
Hi Dave

I agree it’s a interesting question as to how to use it.

Outdoor shots it’s always going to need a 2nd shot to blend back due to movement issues. Problem is the combination as the FA shots appear smoother to my eye i.e. no grain. So combination on a oak tree or pine tree would be a time consuming event vs just running a nose reduction tool.

But I agree it makes sense to run a few in each series.

Still need to get out and do some stream work.

Paul C
 

Craig Stocks

Well-known member
Have you tried varying the number of frames? I wonder how many are necessary to make the noise disappear entirely.
I'd say it depends on your camera, the scene and your expectations. With my test scene and pushing the exposure very aggressively I've found that 10 frames is more than enough with the IQ4150 where the Sony a7R2 needs 16 to 32 frames and my old Sony a6000 never catches up even with 96 frames. (Older Sony cameras support the Play Memories apps which includes one call "Smooth Reflections" that does in-camera averaging and outputs a RAW file.)

There is definitely a trend of diminishing returns as you add additional frames. Each frame contributes 1/n where n is the number of frames. So frame 5 contributes 1/5th, frame 10 contributes 1/10th and frame 100 contributes just 1/100th to the average. Again, considering the a6000, there isn't much difference between 32 and 96 frames.

I think that I will start using frame averaging more and more. I don't see it as much or any of an advantage for a well lit and well exposed scene but it can work wonders if you need to aggressively work a scene post processing. I'll try to use it for foreground frames when doing night sky photos when I can live within the 2 second shutter speed, otherwise I'll continue to stack in Photoshop. We're taking the kids and grand kids to the beach next week and the a6000 will be my only camera so I expect I'll use frame averaging when I can to get more out of an aging camera.

FWIW I wouldn't expect any noise reduction process to be able to compete with frame averaging. Frame averaging truly adds additional data to the image whereas noise reduction can only spread, average or smear existing data.
 

JimKasson

Well-known member
I would posit that "inoffensive" is a better standard than invisible. That is, of course, subjective, but seems to be the final result that you should care about.

Some images benefit from heavy grain. Some do not. The question is what camera, technique, and post processing will get you the image you want with the amount of grain that fits the image (whether high, low, or near zero).
Very sensible. After all, we are presumably photographing with a purpose, and IQ beyond that which achieves our purpose is unnecessary.
 

vjbelle

Well-known member
My thoughts as I have not tried to use Frame Averaging much outdoors is that if there is a chance of 'Any' camera movement due to wind it would probably be advantageous to use a 10 stop ND. Any slight movement would probably not be noticed due to the long exposure where with Frame Averaging (again I haven't tried this under these conditions) any movement is reported to be very destructive.

Random thoughts.......

Victor
 

Craig Stocks

Well-known member
My thoughts as I have not tried to use Frame Averaging much outdoors is that if there is a chance of 'Any' camera movement due to wind it would probably be advantageous to use a 10 stop ND. Any slight movement would probably not be noticed due to the long exposure where with Frame Averaging (again I haven't tried this under these conditions) any movement is reported to be very destructive.

Random thoughts.......

Victor
I think movement is movement regardless of whether you use a 10 stop ND or frame averaging to achieve the same exposure time. Camera movement will probably always be a problem, subject movement may or may not depending on the subject. However, frame averaging with a shutter speed that's too fast for continuous capture may show some artifacts depending on the scene, the motion, the number of frames and the shutter speed. So far in my very limited experience subject motion hasn't been a problem but I can certainly imagine scenarios where it would be.
 

SrMphoto

Well-known member
My thoughts as I have not tried to use Frame Averaging much outdoors is that if there is a chance of 'Any' camera movement due to wind it would probably be advantageous to use a 10 stop ND. Any slight movement would probably not be noticed due to the long exposure where with Frame Averaging (again I haven't tried this under these conditions) any movement is reported to be very destructive.

Random thoughts.......

Victor
Frame averaging (in camera or in post) gives you a similar result as ND filter. Pixel shift is a technology where movement tends to cause problems.
 

vjbelle

Well-known member
I was thinking back to the images presented by Phase with one shot in Skye at the lighthouse. There was some criticism about the slight blur of the image due to some slight camera movement. I am thinking that the slight 'blur' would have been eliminated by using a like time frame with ND's. The camera movement was probably very random and would have been overshadowed/eliminated by the non-movement time.

Again.... I have no experience with this but will in time.

Victor
 

Paul2660

Well-known member
Frame averaging (in camera or in post) gives you a similar result as ND filter. Pixel shift is a technology where movement tends to cause problems.
Movement of any kind is totally blurred in P1 Frame Averagine, even a slight breeze can and will disrupt the shot, and the effect is not the same as with a ND over the same time frame. The blur is most times chopped, even with the frame averaging set to no gapping. Leaves, and tree branches are very easily blurred. And in fact the comment on Pixel shift and movement is a bit misconstrued. Pixel shift at least on a K1 Pentax needed the camera to be still (no different than P1 and Frame Averaging). Subject movement could be handled very well, unless there was a strong wind blowing. LR/ACR did a TERRIBLE job on this, and that was most folks saw. C1 never did a raw conversion for Pixel shift (sadly) but tools like Iridient and or Silky Pix both proved that you can get a very good image from a subject where there was movement. In fact no artifacts at all.

I wrote up a article on my site that shows the differences in what various raw converters did with the K1 images. Sadly, Adobe did their one and done look at the K1 as they do for most raws and C1 never even attempted to convert them. The fact that Iridient could easily out perform Adobe on the raw conversion of K1 Pixel shift was at enhanced after Iridient started to allow Dng output. Prior to that I never used them as their raw tool set can't compare to either C1 or LR/ACR.

https://photosofarkansas.com/2017/06/27/062517-examples-of-pixel-shift-with-motion-correction-between-adobe-lr-and-silkypix/

Frame Averaging in P1 can't handle any subject movement at least from what I have seen so far. You need 100% no wind.

Paul C


Paul C
 

Craig Stocks

Well-known member
It appears there are two basic reasons for using frame averaging, simulating a long exposure like you would get with a neutral density filter or minimizing noise in shadows. Neither use is new but it's exciting to have access to the tool right in camera.

Astrophotographers have been using frame averaging for years to reduce noise in order to brighten and stretch dark areas. This article does a good job of explaining that it's really about total exposure time, not total number of exposures. Astrophotographers frequently talk in terms of the number of hours of total exposure time on a target with all of the frames stacked and averaged.

https://www.skyandtelescope.com/ast...graphy-tips/astrophotography-stacking-signal/
 

Paul2660

Well-known member
Yes, stacking has been in use for quite a while in the Astrophotography world. Mean and Maximum mode in Photoshop, which were as I understand it discovered quite by accident and put into Photoshop around Version 4 or so with the "stack modes".

https://photosofarkansas.com/2014/09/23/092314-using-stacking-for-better-night-photography-results/

I still use this same process when I go after Star Trails, and it can be use for Milky Way work also, but workflow is different, but same result, less noise and better color in the Milky Way.

Paul C
 
Top