Home · Register · Join Upload & Sell

Moderated by: Fred Miranda

New fredmiranda.com Mobile Site

New Feature: SMS Notification alert

New Feature: Buy & Sell Watchlist

FM Forums | Sony Forum | Join Upload & Sell

1 2 3 … 13

15 16 end

Archive 2017 · Pixel shift and lens testing thoughts...

DavidBM
Offline
• • • • •
Upload & Sell: Off

p.14 #1 · p.14 #1 · Pixel shift and lens testing thoughts...

Fred Miranda wrote:
I don't miss the camera apps as they were not a great implementation. I only miss image averaging from the SR app to mimic ND filters and now that's been replaced by real filters.
I'm hopeful image averaging could be added as an in-camera feature to make this camera perfect for me.

On the other hand, although in-camera image averaging is helpful replacing ND filters, the benefit of having improved S/N ratio is much smaller after 5-8 images. Also, too many averaged images tend to create some posterization in solid areas. I see that starting with 32+ averages images. @bclaff_too@ also...Show more →

Do you think the poserization is just due to very low noise (a little noise dithers away poserization) or some other effect from the averaging?

Dec 15, 2017 at 04:46 PM

JimKasson
Offline
• • •
Upload & Sell: Off

p.14 #2 · p.14 #2 · Pixel shift and lens testing thoughts...

DavidBM wrote:
Do you think the posterization is just due to very low noise (a little noise dithers away posterization) or some other effect from the averaging?

Fixed pattern read noise tends to have relatively more low-spatial-frequency components than the frame-to-frame dynamic read noise. Averaging out the latter leaves the former more visible. Maybe that's it.

Jim

Dec 15, 2017 at 05:19 PM

JimKasson
Offline
• • •
Upload & Sell: Off

p.14 #3 · p.14 #3 · Pixel shift and lens testing thoughts...

By the way, here's a comparison of RT pixel-shift processing of the 7-stop underexposed 4 image stack and demosaicing each image using AMaZE and averaging those images.

To my eye, there is more noise and more detail in the pixel-shifted set, which shows the texture of the loth binding of Fan Ho's book.

Averaged after demosaicing

Pixel-shifted

Dec 15, 2017 at 05:23 PM

rico
Online
• • • • •
Upload & Sell: Off

p.14 #4 · p.14 #4 · Pixel shift and lens testing thoughts...

Posterization in this situation is an output limitation, and can be solved with a 10 bit/channel output path. I use a low-end Quadro, DisplayPort cable, and higher-end Eizo monitor. Averaging reduces noise and exposes the smooth gradients of your image. Since my postprocessing includes the Gimp (8 bit/channel only), these gradients are immediately evident if I run my output in 8 bits/channel. For web pics, I find it necessary to add noise in certain cases. Back when video cards showed 256 colors only (sometimes backed by a color LUT), dithering algorithms (introduced noise) were an active area of research.

Dec 15, 2017 at 06:13 PM

Fred Miranda
Offline
Admin
Upload & Sell: On

p.14 #5 · p.14 #5 · Pixel shift and lens testing thoughts...

DavidBM wrote:
Do you think the poserization is just due to very low noise (a little noise dithers away poserization) or some other effect from the averaging?

I believe it's due to "Fixed Pattern Noise" which is worse on the A7R II. Bill Claff posted a test here but I can't find the link. What he found on his test is visible in real world images (solid areas like sky, still water, etc.) when averaging too many images via the SR app (I'd say more than 32 images).
It looks like you are correct when you say there is little noise to dither the fixed pattern and therefore it becomes evident.

Dec 15, 2017 at 06:30 PM

tn1krr
Offline
• •
Upload & Sell: Off

p.14 #6 · p.14 #6 · Pixel shift and lens testing thoughts...

Fred Miranda wrote:
I believe it's due to "Fixed Pattern Noise" which is worse on the A7R II. Bill Claff posted a test here but I can't find the link. What he found on his test is visible in real world images (solid areas like sky, still water, etc.) when averaging too many images via the SR app (I'd say more than 32 images).
It looks like you are correct when you say there is little noise to dither the fixed pattern and therefore it becomes evident.

Pretty much what this comparison from Bill's site shows? A7RIII is way more even in terms of sensor self-heating.

http://www.photonstophotos.net/Charts/Sensor_Heatmaps.htm#mode=23,camera=,suffix=,cameraC=Sony%20ILCE-7RM2,suffixC=14

Dec 15, 2017 at 06:39 PM

Fred Miranda
Offline
Admin
Upload & Sell: On

p.14 #7 · p.14 #7 · Pixel shift and lens testing thoughts...

tn1krr wrote:
Pretty much what this comparison from Bill's site shows? A7RIII is way more even in terms of sensor self-heating.

http://www.photonstophotos.net/Charts/Sensor_Heatmaps.htm#mode=23,camera=,suffix=,cameraC=Sony%20ILCE-7RM2,suffixC=14

Perfect!

Dec 15, 2017 at 06:41 PM

lightskyland
Offline
• • •
Upload & Sell: On

p.14 #8 · p.14 #8 · Pixel shift and lens testing thoughts...

Luvwine wrote:
Would like for that to be true. I have not yet tried it, but so far have not seen many examples of successful out of doors and non architecture use of PS. I do hope this will be useful for calm morning landscapes and such, but am concerned that it may be only rarely useful. Time will tell, I guess.

I've taken some at Jordan Lake but haven't processed them yet. I need to get around to that. Been very busy with my business so haven't had time for the "deep dive" yet to adapt my workflow to pixel shift. I figure it will be useful for details such as rocks and distant features, and not useful for any foliage. So just mask the latter with a single shot from the pixel shift.

Dec 15, 2017 at 07:22 PM

Fred Miranda
Offline
Admin
Upload & Sell: On

p.14 #9 · p.14 #9 · Pixel shift and lens testing thoughts...

Initially I've read that the minimum pixel shift delay between shots was 0.5s (perhaps with the pre-production model).
The production mode has a minimum of 1 second delay and that's huge with much higher susceptibility to movement, including camera shake, wind. Half a second was already too long imo.

In comparison Pentax's pixel shift implementation has pretty much zero lag between shots and uses the same technique...

Can we be hopeful this delay could be lowered in a future firmware?

Dec 16, 2017 at 01:23 AM

mttran
Offline
• • • • •
Upload & Sell: Off

p.14 #10 · p.14 #10 · Pixel shift and lens testing thoughts...

Luminous Landscape A7R iii pixel shift and Phase One Trichromatic comparison files: https://luminous-landscape.com/sony-a7r-iii-pixel-shift/

Dec 16, 2017 at 07:15 PM

JimKasson
Offline
• • •
Upload & Sell: Off

p.14 #11 · p.14 #11 · Pixel shift and lens testing thoughts...

mttran wrote:
Luminous Landscape A7R iii pixel shift and Phase One Trichromatic comparison files: https://luminous-landscape.com/sony-a7r-iii-pixel-shift/

I don't find this test to be very useful because of the extreme sharpening introduced by Imaging Edge. To be fair, alternatives for processing ARQ files have only recently become available.

Jim

Dec 17, 2017 at 01:01 PM

GMPhotography
Offline
• • • • • • •
Upload & Sell: On

p.14 #12 · p.14 #12 · Pixel shift and lens testing thoughts...

I’m dumping Image Edge ASAP. I’m a C1 user so I can take a Tif back in at worst case and edit more. Just need to see which software will work

Dec 17, 2017 at 01:20 PM

Fred Miranda
Offline
Admin
Upload & Sell: On

p.14 #13 · p.14 #13 · Pixel shift and lens testing thoughts...

JimKasson wrote:
I don't find this test to be very useful because of the extreme sharpening introduced by Imaging Edge. To be fair, alternatives for processing ARQ files have only recently become available.

Jim

My thoughts as well.

Imaging Edge's sharpening is aggressive. I've seen some folks comparing a pixel shift composite from Imaging Edge vs a regular LR's RAW and coming to the conclusion that the pixel shift result is amazing.
From my own tests, using Iridient and RT, the difference in 'resolution' is mostly due to different sharpening recipes. There is indeed slight more high frequency detail in the composite image but although the difference is obvious at 1:1, it’s smaller than Sony and other sites are reporting.

If the delay between images was minimal (Like Pentax's implementation), I would perhaps use it for landscapes but it takes more than 4 seconds for a complete sequence and therefore subject motion is pretty much guaranteed, and therefore artifacts will be a big problem to deal with in post.

Dec 17, 2017 at 05:14 PM

mttran
Offline
• • • • •
Upload & Sell: Off

p.14 #14 · p.14 #14 · Pixel shift and lens testing thoughts...

JimKasson wrote:
I don't find this test to be very useful because of the extreme sharpening introduced by Imaging Edge. To be fair, alternatives for processing ARQ files have only recently become available.

Jim

+1, I play some more at home with Guy's bronco in ARQ format @ the most conservative Imaging Edge setting and its TIFF result seem not too bad after PS. Yes, the luminous-landscape pixel shifted first trial out samples are not that good as you mentioned but Imaging Edge pixel shift process can still be easily managed sharpness in both ARQ and ARW formats.

Dec 17, 2017 at 05:37 PM

lightskyland
Offline
• • •
Upload & Sell: On

p.14 #15 · p.14 #15 · Pixel shift and lens testing thoughts...

To me, the resolution difference for detail in the Nyquist regime is very apparent and obvious. Particularly for subjects with significant coloration.

I am basing this on Jim Kasson's samples, Guy's Big Bronco images (the wood grain in particular) as well as those at DPReview.

I'm certainly going to continue to use this in my own tripod-based captures for maximum detail.

Dec 17, 2017 at 07:20 PM

GMPhotography
Offline
• • • • • • •
Upload & Sell: On

p.14 #16 · p.14 #16 · Pixel shift and lens testing thoughts...

Good to see you folks using the Big Bronco shots to look at this tech. I think it has a great usage advantage when it makes sense. But it’s not going to appeal to too many type of shots. I look at it as a feature set that maybe useful to take any advantage over a single shot for maximum draw you can get out of your glass and treading on Medium formats door. But it’s gotta work and we need stellar software for it. Now we just need to see who will come out on top with software use. Image Edge is not it, it’s simply not Professional software. That’s okay I get this is not Sony’s area of expertise. No harm no foul. Unlike some reviewers total bashing comments of Sony’s which is really getting sickening to me. But I will say having a partnership with Phase One they would have been better well served giving them the ball to run with this tech. I’m actually baffled why they did not when C1 is free to Sony users.

Dec 17, 2017 at 08:29 PM

DavidBM
Offline
• • • • •
Upload & Sell: Off

p.14 #17 · p.14 #17 · Pixel shift and lens testing thoughts...

Fred Miranda wrote:
My thoughts as well.

Imaging Edge's sharpening is aggressive. I've seen some folks comparing a pixel shift composite from Imaging Edge vs a regular LR's RAW and coming to the conclusion that the pixel shift result is amazing.
From my own tests, using Iridient and RT, the difference in 'resolution' is mostly due to different sharpening recipes. There is indeed slight more high frequency detail in the composite image but although the difference is obvious at 1:1, it’s smaller than Sony and other sites are reporting.

If the delay between images was minimal (Like Pentax's implementation), I would perhaps use it...Show more →

Yes a lot of what we are seeing is down to aggressive sharpening. But then it does seem to me that these files can take more aggressive sharpening before displaying artefacts than regular ones, and if that's so then it's a kind of win.

Dec 17, 2017 at 10:09 PM

Fred Miranda
Offline
Admin
Upload & Sell: On

p.14 #18 · p.14 #18 · Pixel shift and lens testing thoughts...

DavidBM wrote:
Yes a lot of what we are seeing is down to aggressive sharpening. But then it does seem to me that these files can take more aggressive sharpening before displaying artefacts than regular ones, and if that's so then it's a kind of win.

True but I think what we are seeing is a combination of things:

1) No demosaicing means we are not sharpening (amplifying) 'errors'.

2) Image averaging when combining the pixel shift sequence give us more than a stop higher DR and improved S/N ratio which means we are sharpening less noise as well...

That thing is, image averaging could be done without pixel shift...

Like I wrote, there are applications that will benefit tremendously from not having false color and aliasing but I'm not sure landscapes is one of them. I do see an increase in resolution which could be useful but Sony's implementation with a huge time delay between shots kills it for me.

Dec 17, 2017 at 10:16 PM

GMPhotography
Offline
• • • • • • •
Upload & Sell: On

p.14 #19 · p.14 #19 · Pixel shift and lens testing thoughts...

Exactly my thought David. End of day it’s the file if we can be aggressive to a point we don’t start destroying the file than it’s a win. Really what we need is someone with a Epson 990O printer to go large on a standard image and one of these. My bet is some of the agreessive look won’t show up but still be highly detailed.

I have yet to try Iridient so I’m leaving them out of this comment. We just need a more mature Professional software to handle these files correctly. Once we have that this could be really special

Dec 17, 2017 at 10:16 PM

GMPhotography
Offline
• • • • • • •
Upload & Sell: On

p.14 #20 · p.14 #20 · Pixel shift and lens testing thoughts...

Not sure you can do this and certainly worth testing this out. If you can take one of the 4 files and actually go negative on the sharpening than take those settings apply it to the other 3 than do the pixel stitch it may offset that algorithm that they are using on the stitch. This maybe worth a try. Than do any sharpening you want in PSor raw converter we like.

These are the times I regret selling my Epson 7900 printer. Loved that thing

Dec 17, 2017 at 10:20 PM

1 2 3 … 13 14 15 16 end
FM Forums \| Sony Forum \| Join Upload & Sell 1 2 3 … 13 14 15 16 end

This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.