While I find the term "FF image quality" a bit misleading it is quite an interesting use for the high burst rate.
It should be fairly easy to fully automate the workflow with a bit of OpenCV programming. The software could automatically determine corresponding images from the same bursts on the SD card by calculating an image similarity score (raw conversion would be performed with DCraw). Each burst is then auto-aligned and the mean or the median for each pixel is calculated.
I think this sounds like a lot of files, but not all that much extra work (sounds like less than typical HDR processing). The point it that if you want the ability to travel really light and still get some high-value results, this allows you to do it. I'd can see working a little harder at my comfy desk for that benefit.
I used to use stacking to get more DR out of my 5D3's sensor. Shoot at iso 400 to kill the banding then stack ten shots to kill the noise and get smooth tonality. Now I get a far superior file in one frame from my A7r.
Mark_L wrote:
Have you also tried PhotoAcute and similar for more resolution?
I just checked the pa website, don't see any profiles for the Nikon 1 system. I've found that PA lives up to expectations only with a matching profile.
Having said that, one can do up the profile images and send them in. The next release of PA will contain the new profile(s), and this earns a free PA licence The next release can take months or longer.... but with enough patience it works.
M635_Guy wrote:
I think this sounds like a lot of files, but not all that much extra work (sounds like less than typical HDR processing). The point it that if you want the ability to travel really light and still get some high-value results, this allows you to do it. I'd can see working a little harder at my comfy desk for that benefit.
typical HDR would need only 3 to 7 images. This definitely needs a lot more (stop of noise reduction = (# pics)^0.5). The processing time is not trivial, especially in PhotoAcute.
Yep, the PhotoAcute sample is crunchy because I sharpened it using the same workflow as the others to keep it comparable and there are some artifacts sharpening the PhotoAcute image. PhotoAcute doesn't have any Nikon 1 profiles so I was forced to randomly choose a D800E instead; agreed that using native profiles (if they were available) would achieve better results.
Here is Sample #3, this time at High ISO (3200). For High ISO, ETTR factors less because there is less DR to work with and because you're shutter-speed limited (otherwise you wouldn't need to use a higher ISO). That said, you generally want to use the Highest ISO you can without clipping the exposure, using the same absolute exposure (ie aperture and shutter) that you would have used at a lower ISO, because the Nikon 1 sensors have less read noise at Higher ISOs. Like the previous two samples, these were all taken hand-held. All are downsampled to 8MP.
multi-frame noise reduction is already implemented in some Sony cameras, though obviously they only give out a jpeg file. Also they only do it at very high ISO to extend beyond the normal ISO range.
curious80 wrote:
multi-frame noise reduction is already implemented in some Sony cameras, though obviously they only give out a jpeg file. Also they only do it at very high ISO to extend beyond the normal ISO range.
Plus they shoot at a much slower burst rate - the J4 can shoot 20 frames in 1/3 second, whereas the Sony cameras require 1 second to shoot approx 10 frames.
snapsy wrote:
Yep, the PhotoAcute sample is crunchy because I sharpened it using the same workflow as the others to keep it comparable and there are some artifacts sharpening the PhotoAcute image. PhotoAcute doesn't have any Nikon 1 profiles so I was forced to randomly choose a D800E instead; agreed that using native profiles (if they were available) would achieve better results.
Do you think you are getting better, worse or similar results from ACR vs. PA?
galenapass wrote:
Do you think you are getting better, worse or similar results from ACR vs. PA?
PA yielded the same noise benefits of ACR with the added benefit of higher acuity. Unfortunately it doesn't look like PA will ever add profile support for the Nikon 1 because they've stopped development of the software (link).
kaplah wrote:
! That's bad news. It's the only good superresolution software out there.
Ohh so that's why they don't have a profile for my A6000 either. I tried it out last month when I was in search of a super-resolution software. I ended up having to use NEX-6 profile which still gave pretty good result for my test shots. However I then tried it on a burst of moon shots but wasn't really able to extract any real resolution increase out of it.
snapsy wrote:
Plus they shoot at a much slower burst rate - the J4 can shoot 20 frames in 1/3 second, whereas the Sony cameras require 1 second to shoot approx 10 frames.
If you are shooting a static scene of the type where stacking is likely to work well, then I don't think having the burst complete in 1 second should cause much of an issue.
if you were out & only brought a small sensor camera with you, this seems like a good way to preserve an important image
otherwise, it seems like a LOT of work -- I hope nikon will develop this system as it would be great to have something to stick on a long Nikon tele for reach
curious80 wrote:
If you are shooting a static scene of the type where stacking is likely to work well, then I don't think having the burst complete in 1 second should cause much of an issue.
It would take the RX100 2 seconds to burst 20 photos, 6x the amount of time as the J4/V3. It's doable, just prone to more inter-frame movement.
Not sure if I read that article. I think the first place I read about the technique was on dpreview many years ago. Thom Hogan has an article about it using the Nikon 1 series bodies as well.
There are a lot of articles about stacking - it's been around as long as digital imaging. I imagine there's a lot of recent interest from the E-M5 II's built-in superresolution mode.
The next release can take months or longer.... but with enough patience it works.