Wayne Fox
Workshop Member
i think with 20 shots you’ll still see blurry streaks. i think to get rid of people you would be better off taking a series manually and averaging in post.
Might work if the tool allowed a delay between captures.
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!
Phase One IQ4 - Feature Update 1
- Thread starter dougpeterson
- Start date
Might work if the tool allowed a delay between captures.
dougpeterson
Workshop Member
Disappearing people is better done with median than mean.
I hope (and will be advocating for) the addition of median to the tool in the future. I'd also like to see max, which is very useful for light painting and some kinds of astronomical images.
dougpeterson
Workshop Member
Last week it was "expected next week". I've not had an update since then, but we're still in the early part of the week. Hopefully they don't push it out until it's thoroughly tested (the beta process for the dealers gives me strong reason to believe that this is the case); the vertical bug on the last firmware update was, frankly, embarrassing.
Doug
Not trying to shoot the messenger, Doug, but P1 does have a bit of cause for concern. 7 months and still no working auto rotation.
What IMO is embarrassing is that from day 1 the auto rotation did not work. As you well know. If you captured anything in the vertical orientation, once the back was put back to horizontal, all the vertical images also displayed horizontal, which made playback next to impossible. Rotation of a back on the XF, with an L bracket no big deal, rotation of a back on a tech camera, any tech I know of means, releasing the back from it's current position, which exposes the back to the elements, then rotating the back on tech mount to vertical, and of course the opposite is true when coming back to horizontal. Auto rotation well I guess I just consider that digital camera 101, pretty basic requirement, period.
The fact that the "fix" for auto rotation brought out the multiple image blend error on playback (for what I have been told was all verticals) well, I leave that for the jury, but amazing it was not caught.
BTW, there still seems to be no word on the multiple image split on playback issue as both of the IQ4 backs I have used will still do this at times. Still not on the "problem" firmware that created the mess for verticals but assuming it did not address the issue for horizontals since the problem just become more problematic for all verticals.
The release of fixes seems to be taking way too long on this particular back. Beta has been out a week, fixes needed should have been back to P1 in a day, and I am pretty sure the actual beta was floating around longer than last week, as it seems now that several of the "chosen" P1 professionals have had for a longer.
Paul C
MGrayson
Subscriber and Workshop Member
Doug,
I'm not up on the latest algorithms, but a running mean can be computed storing only one extra frame of data and the number of frames so far- the new mean is a weighted average of the old mean and the new data. To compute a median, I think you'd have to store all the data from every frame and, as you mentioned earlier, that gets to tens of gigabytes quickly. Perhaps you could combine the two and take the median of a bunch of averages? I dunno, but it's not as easy as just taking the mean.
Best,
Matt
(In fact, I think this is the only reason the mean is a popular "middle" estimator. It's terribly non-robust. One outlier ruins it. But the means of two data sets can be computed without knowing anything but the means and number of elements in each. Other, more robust estimators, like median, are better in every way...except that you can't easily compute the median of the union of two data sets - it's nonlinear.)
dougpeterson
Workshop Member
Given that we're doing photography and not mission-critical statistics I suspect a good solution can be had within the constraint of 10-20 frames worth of data.
For example:
1) Capture 10 images, compute and save a median of those frames, discard the data of the 10 captures themselves. This requires 11 frames of memory along the way and ends with 1 frame of data.
2) Capture another set of 10 images, compute and save a median of that (second) set. This requires 12 frames of memory (11 for this step, plus remembering the median-of-first-10) and ends with 2 frames of data.
3) Capture another set of 10 images, compute and save a median of that (third) set. This requires 13 frames of memory (11 for this step, plus the median-of-first-10 and median-of-second-10) and ends with 3 frames of data.
4-10) Rinse and repeat. On the 10th go around you need a total of 21 frames of memory and end with 10 frames of data, each representing the median of ten images.
11) Take the median of the 10 set-medians.
This won't, strictly speaking, produce the same median result as a median of all 100 frames. But it should both greatly improve noise and get rid of people/things that are moving in the frame unless they are there for more than half the total frames.
If desired...
12) Store the results of (11) as the median-of-first-100 and rinse repeat 9 more times such that you have a (more-or-less) median frame for 1,000 frames, which only requires 21 frames of storage at any given moment.
13) Do another level of recursion for a total of 10,000 frames that only requires 21 frames of storage at any given moment.
Of course you could do this with any number, not just 10, and I assume it would be smarter to use an odd number.
Math is not my forte, and I didn't get much sleep last night, so there could be any number of mistakes above. But the general idea should be right.
dougpeterson
Workshop Member
I agree with your concern. Hopefully this Feature Update is the start of a real turnaround in this regard!
Audii-Dudii
Active member
I'm no expert in these matters, to be sure, but I've been blending multiple files into one to enhance image quality for many years now and many hundreds, if not thousands of photos, and I agree that a fairly small number of frames is sufficient for most purposes.
To reduce noise, increase SNR, and eliminate motion effects in a photo -- the primary benefits to be achieved by blending files for a typical stills photographer -- it's not necessary to blend hundreds or thousands of files into one.
That's only necessary for astronomers, who aren't trying to reduce noise or increase the effective SNR of their camera's sensor, as much as they are trying increase the total amount of light that is gathered when taking a photo, which is another matter altogether.
As usually is the case with most things, at some point, as the number of frames being blended increases, the benefits that result from blending them will begin to diminish with each additional frame.
(As I noted in another thread, I find seven files to be a good compromise for my long-exposure nighttime photos. That's because standing around, twiddling my thumbs while waiting on my camera for any longer than,say, 5 to 7 minutes, begins to reduce the fun of photography for me, even if doing so might result in a very slight improvement in the quality of the final, finished photos. Of course, this will be much less of an issue when using significantly shorter exposures for daytime photography, but everyone will need to determine where to draw that line for themselves...)
For some modes of blending, such as mean or averaging, it doesn't matter how many files are blended or whether the number of files is even or odd. However, because the median value of a sorted list of numbers is the middle value, when blending in median mode, blending an odd number of files is preferred.
Otherwise, the software will have to calculate the median value for each pixel by averaging the middle two values and this potentially means every pixel value of the final, blended file could, in effect, be made up instead of being selected from among the range of real values that were actually recorded. While this may not matter very much in practice, in theory, at least, blending an odd number of files is preferred, if only to preserve the already tenuous connection between the pixels captured by the camera and those used to create the finished photo.
dougpeterson
Workshop Member
Feature Update 1 is now live!
https://www.phaseone.com/en/SupportMain/Camera-Firmware/IQ
Our support team is working until 6:30pm ET today in the case any of our clients wish to update and have questions.
A reasonable warning: any firmware update of any device (phone, camera, computer etc) carries a very small risk of “bricking” that device. Of course every IQ4 comes with a five year warranty that includes a loaner during any repair or service, but with the holiday weekend upon us I would discourage anyone from updating if they have an important production/job/shoot this weekend. Or if you update, do it early today so we have time to assist you in the (rare) chance of any issue.
Happy shooting everyone!
For those in the USA I wonder how frame averaging will handle fireworks??
Nice of P1 to celebrate the US Independence Day!
dougpeterson
Workshop Member
Our support team has created the following guide to installing the new firmware:
https://www.dtcommercialphoto.com/updating-your-iq4-with-the-creative-control-firmware/
https://www.dtcommercialphoto.com/updating-your-iq4-with-the-creative-control-firmware/
If it works as intended it will cancel the fireworks out. So that is one situation where you would want a long exposure.
What might cause an interesting effect is doing a long exposure + frame averaging. It would in essence capture all the fireworks at once maybe?
Cancelling fireworks out (or removing people) is done with median averaging. Phase One’s method uses mean averaging which produces the same effect as ND filters.
dougpeterson
Workshop Member
Update: the max frame and maximum duration were both raised for the production version of the firmware. With an IQ4 150mp the max frame count is 3600 and the maximum total duration is 2 hours. This is a major improvement from the original 1000 count and 12 minutes originally expected.
Bill Caulfeild-Browne
Well-known member
I have successfully installed the update, but I confess I don't have a clue how to best set up and utilize the Frame Averaging! Is there an instruction sheet or demo video - or do I just experiment? Like, if you're trying to reduce noise in a landscape image, what would you set?
The aperture and exposure time are presumably what you'd use without FA, but how do you determine how many exposures?
Looks like fun but I need to learn
Cheers,
Bill
The aperture and exposure time are presumably what you'd use without FA, but how do you determine how many exposures?
Looks like fun but I need to learn
Cheers,
Bill
Hi Bill
Best way I have determined is you set your desired shutter speed then pick the time interval that would wish.
So say 1/200 is the shutter 0.9 seconds your interval. The number of exposures would be 0.9 seconds divided by 1/200 of a sec.
You also need to pick an ISO and aperture.
Tool can only be fired currently from the shutter button on LCD so remote release on XF won’t trigger it. You get an exposure just not a averaged one. When tool run you get a small status bar bottom right.
With XF set a timer delay around 3 to 5 sec to help keep XF vibration reduced since took doesn’t respect the XF vibration reduction seismic tool.
Subject need to have really no movement even slight wind will create blur.
So Limited use for me. Out west shooting a mountain range this would be wonderful. I also think it might reduce the radiant heat blurring but would have to play with it. Longer your interval the worse the motion issues become.
But the difference in noise for shadows is impressive so it a tool I will try to get a base exposure with.
Blending back due to the amount of blur can become very problematic depending on subject.
Paul C
Best way I have determined is you set your desired shutter speed then pick the time interval that would wish.
So say 1/200 is the shutter 0.9 seconds your interval. The number of exposures would be 0.9 seconds divided by 1/200 of a sec.
You also need to pick an ISO and aperture.
Tool can only be fired currently from the shutter button on LCD so remote release on XF won’t trigger it. You get an exposure just not a averaged one. When tool run you get a small status bar bottom right.
With XF set a timer delay around 3 to 5 sec to help keep XF vibration reduced since took doesn’t respect the XF vibration reduction seismic tool.
Subject need to have really no movement even slight wind will create blur.
So Limited use for me. Out west shooting a mountain range this would be wonderful. I also think it might reduce the radiant heat blurring but would have to play with it. Longer your interval the worse the motion issues become.
But the difference in noise for shadows is impressive so it a tool I will try to get a base exposure with.
Blending back due to the amount of blur can become very problematic depending on subject.
Paul C
Bill Caulfeild-Browne
Well-known member
Thanks Paul, that's very helpful. Gives me a base for experimenting. Hope to show some results soon.
Bill
Bill