D800

  • Thread starter Thread starter Guest
  • Start date Start date
dfreerider said:
True, sharpening can't create detail that is not there to begin with. However it is a visual trick of enhancing edge contrast, and as such sharpening in post in still photography is usually needed to some extent. How much and how depends on many factors from how the photo was shot (camera make and settings, raw vs jpg etc.). All jpgs coming out of the camera will have digital sharpening applied to some extent. Raws are usually softer out of the camera, and resizing will make them even softer. Finally your final format will require different sharpening, print often requires more sharpening than the computer screen etc.

When it comes to DSLR video with all its limitations, the sharpening algorithms are often very crude due to data processing restraints, so shooting soft and post sharpening it is often the better way to go.
Click to expand...

Agreed. Though i do think a lot of people over sharpen their images and IMHO Mr Bloom made the same mistake with Mk3.
 
I don't care about the moire or aliasing all that much. It's the crazy noise that some tests show that has me very concerned. For example:

http://www.youtube.com/watch?v=ey8el-YcvgA&feature=related
http://vimeo.com/39326666
http://vimeo.com/album/1884940/video/39326883
http://vimeo.com/album/1884940/video/39327142

You can even see it if you look closely at the arm of the black chair in this shot:
http://vimeo.com/album/1884940/video/39326376

But in other examples, I don't see this at all. For example:
http://vimeo.com/38912704
http://vimeo.com/39120089

Very odd.
 
Last edited:
latyshev said:
USM simply cannot bring out detail which is not there. And detail cannot "somehow be encoded in the blur". If its blurred, its blurred. It makes an illusion of a sharp image, but these are two different things.

Ive been experimenting with video for only a year, but i've had my fair share of still work. And USM just makes images look fake. But there are people who like it and sharpen the hell out of their pictures.

Compare a wall sized print from 6mp with sharpening and a 16mp print with 0 sharpening. Same thing with video... Mk3 vs C300.

Another analogy (taking it to the extreme) is applying DOF in post... yikes.
Click to expand...
latyshev said:
(...) i do think a lot of people over sharpen their images and IMHO Mr Bloom made the same mistake with Mk3.
Click to expand...
I've followed your posts on dvxuser and outside. I was almost to answer you before. Now I will.

Your conservative overview can be both: negative and positive (Believe it! Canon DSLR 4K will be released earlier than your pessimist time frame :-)

(on topic now) The proof is in the pudding! -- never forget that one.

With artifacts ala 5DII is impossible to apply certain tools. Not with a clean output like 5DIII offers. This is what this camera has brought to us. Period. Bob is right.

You are right in one thing though. Not that the detail is not there. It is, but hidden in that much softness. However, with post processing, you are pushing it to the limits. What will reforce a digital look. Optical DOF (your digital analogy is superb, my hat off ;) can balance.

Then, I dare myself to affirm. Now you can mimic RED output. RED is original. Because happens in-camera. For some reason their cameras cost much more ;-)
 
Emanuel said:
I've followed your posts on dvxuser and outside. I was almost to answer you before. Now I will.

Your conservative overview can be both: negative and positive (Believe it! Canon DSLR 4K will be released earlier than your pessimist time frame :-)
Click to expand...

Is that supposed to be a good or a bad thing? :engel017:

At least i try to base my opinion on some logic and adjust it based on what people say. Eg i can see 4k dslr happening now based on mjpeg, but i still don't think that it makes much sense.

Emanuel said:
(on topic now) The proof is in the pudding! -- never forget that one.

With artifacts ala 5DII is impossible to apply certain tools. Not with a clean output like 5DIII offers. This is what this camera has brought to us. Period. Bob is right.

You are right in one thing though. Not that the detail is not there. It is, but hidden in that much softness. However, with post processing, you are pushing it to the limits. What will reforce a digital look. Optical DOF (your digital analogy is superb, my hat off ;) can balance.

Then, I dare myself to affirm. Now you can mimic RED output. RED is original. Because happens in-camera. For some reason their cameras cost much more ;-)
Click to expand...

"Her favorite past time is eating cookies and her favorite sport is jumping to conclusions" :grin:

I personally think its too early to say. Blurring the hell out of picture to avoid artifacts (and apply aggressive NR at high iso) and then sharpening it in post is a bit like scratching left ear with the right hand. Mr Bloom and many others sing praises, but i don't. To my unproffesional (in video world) eye, the footage from Mk3 just doesn't look that good. Certainly not RED good. And i am not impressed with the codec either. In fact i think that in majority of cases footage from D800 looks almost distinguishably better, and that is a HUGE step for Nikon. I don't see any big steps from Canon. Just IMHO.

Zacuto shootout 2012 will really show who's who ;-)
 
BobbyMurcerFan said:
I don't care about the moire or aliasing all that much. It's the crazy noise that some tests show that has me very concerned. For example:

http://www.youtube.com/watch?v=ey8el-YcvgA&feature=related
http://vimeo.com/39326666
http://vimeo.com/album/1884940/video/39326883
http://vimeo.com/album/1884940/video/39327142

You can even see it if you look closely at the arm of the black chair in this shot:
http://vimeo.com/album/1884940/video/39326376

But in other examples, I don't see this at all. For example:
http://vimeo.com/38912704
http://vimeo.com/39120089

Very odd.
Click to expand...

Check out a quicky i did yesterday https://vimeo.com/39334369 (download in prores). Last scene is iso3200.

I don't think noise pattern can be appreciated on vimeo output - hence the download. Notice that colors are specifically red-orange heavy. I'd say the camera is perfectly usable upto iso1600. BUT since the noise pattern is fine and detail is there, NR techniques like NeatVideo can probably do wonders.
 
I think USM can't bring back detail that's not there; it can make it more visible, but if you overdo it then artifacts will appear, so you have to apply with care - now, if I understand it well (and it may not be the case) convolution sharpness is a completely different beast, it tries to undo what an antialiasing filter has done; if it is actually the case, knowing softness appeared thanks to a software AA filter is a great advantage, it allows the filter to be a lot more aggressive without bringing up artifacts, and bring up, instead, detail that IS there but that your eyes can't see ("given this footage was treated with a strong AA filter, this subtle but very particular pattern I'm seeing can only appear if the original image looked somewhat like this") - or maybe I'm just being an overoptimistic fool, as usual
 
latyshev,

I clicked the link but got an error message. Also got an error message when I tried to download the file. Thanks for the help.
 
Samuel H said:
I think USM can't bring back detail that's not there; it can make it more visible, but if you overdo it then artifacts will appear, so you have to apply with care - now, if I understand it well (and it may not be the case) convolution sharpness is a completely different beast, it tries to undo what an antialiasing filter has done; if it is actually the case, knowing softness appeared thanks to a software AA filter is a great advantage, it allows the filter to be a lot more aggressive without bringing up artifacts, and bring up, instead, detail that IS there but that your eyes can't see ("given this footage was treated with a strong AA filter, this subtle but very particular pattern I'm seeing can only appear if the original image looked somewhat like this") - or maybe I'm just being an overoptimistic fool, as usual
Click to expand...

As far as i know unconvolution technique is based on Fourier series analysis. And as such cannot restore the data fully, but can make a best guess. I was studying this technique way back in university and it can produce some impressive results when the blur is uniform. But everything gets a lot more complicated in real world... Eg - how can you tell OOF vs blurred? Etc...
 
latyshev said:
Is that supposed to be a good or a bad thing? :engel017:

At least i try to base my opinion on some logic and adjust it based on what people say. Eg i can see 4k dslr happening now based on mjpeg, but i still don't think that it makes much sense.



"Her favorite past time is eating cookies and her favorite sport is jumping to conclusions" :grin:

I personally think its too early to say. Blurring the hell out of picture to avoid artifacts (and apply aggressive NR at high iso) and then sharpening it in post is a bit like scratching left ear with the right hand. Mr Bloom and many others sing praises, but i don't. To my unproffesional (in video world) eye, the footage from Mk3 just doesn't look that good. Certainly not RED good. And i am not impressed with the codec either. In fact i think that in majority of cases footage from D800 looks almost distinguishably better, and that is a HUGE step for Nikon. I don't see any big steps from Canon. Just IMHO.

Zacuto shootout 2012 will really show who's who ;-)
Click to expand...
LOL

Well, RED is RAW and the philosophy is going on post anyway.

Nikon was practically in 0 or nearby ;) compared with the competitors. These are their first motion picture cameras ;-) D800 is still to rule out waiting for the 3rd party mosaic engineering's anti-aliasing filter.

Canon 5DII is along F900, DVX (because of dvxuser :-), RED ONE and GH series, one of the most important digital cameras ever launched, if not the best seller. Like happens with the opinions, everyone has one : D

They have C300, C-DSLR 4K to launch, why bother unless to solve the aliasing, moiré and lack of audio support?
NR high ISOs, higher bitrate and new codec are already a bonus.


EDIT -- Last but... Zacuto 2012 shootout is already done (last February). The idea was to have the new cameras if the manufacturers had provided them just in time.
 
Last edited:
Samuel H said:
I think USM can't bring back detail that's not there; it can make it more visible, but if you overdo it then artifacts will appear, so you have to apply with care - now, if I understand it well (and it may not be the case) convolution sharpness is a completely different beast, it tries to undo what an antialiasing filter has done; if it is actually the case, knowing softness appeared thanks to a software AA filter is a great advantage, it allows the filter to be a lot more aggressive without bringing up artifacts, and bring up, instead, detail that IS there but that your eyes can't see ("given this footage was treated with a strong AA filter, this subtle but very particular pattern I'm seeing can only appear if the original image looked somewhat like this") - or maybe I'm just being an overoptimistic fool, as usual
Click to expand...
Not fool at all :-) you are just on the spot, whether we like the digital look whether not.
 
(I would definitely cross GH series of that list.)

I'm just going by what i see so far. "Bluring the hell out" does not equate to "solving" imho. Neither is writing down iso ranges that have nothing to do with actual performance and neither is writing down bit rates that dont correspond to quality improvements.

Anyways, as you said, the proof will be in the pudding :-)
 
latyshev said:
(I would definitely cross GH series of that list.)

I'm just going by what i see so far. "Bluring the hell out" does not equate to "solving" imho. Neither is writing down iso ranges that have nothing to do with actual performance and neither is writing down bit rates that dont correspond to quality improvements.

Anyways, as you said, the proof will be in the pudding :-)
Click to expand...
;))

You mean, hacked GH series? ;)

Right. their 'solution' is to justify the purchase of their next release. 5DIII with workarounds can be what 5DII is not now because with her flaws 'she' is already obsolete with the new girl(s ;-)

 
latyshev said:
Vimeo dropped it after a day... redownloaded now, should work. Same link
https://vimeo.com/39334369
Click to expand...
still doesnt work , error loading message

i join the opnion that USM is not bringing up new detail ,it might look like it does but its just creates the illusion it does
, and from my own experience it looks like that because it accentuate the highlights and bring them out and enhancing contrast also to the blacks, and if pushed too far its creating a kind of strange halo around objects , i dont think it is not legit to use all kinds of post processing to footage , but with all respect , Phillip bloom saying it brings up new detail hiding in the compression/footage - is a complete BS , i know what he means but still from technical point of view wrong , from a perceptive point of view , he's right

its explained rather well in wikipedia

http://en.wikipedia.org/wiki/Unsharp_masking

anyhow found two more vimeo videos of the D4
https://vimeo.com/39402210
https://vimeo.com/39397190

and one more of the D800
https://vimeo.com/39409276
 
Last edited:
we need someone to shoot a resolution chart with 5D2 and 5D3, and apply different sharpening methods...
 
t.p. said:
(...)and if pushed too far its creating a kind of strange halo around objects , i dont think it is not legit to use all kinds of post processing to footage , but with all respect , Phillip bloom saying it brings up new detail hiding in the compression/footage - is a complete BS , i know what he means but still from technical point of view wrong , from a perceptive point of view , he's right

its explained rather well in wikipedia

http://en.wikipedia.org/wiki/Unsharp_masking
Click to expand...
Perception as have been posted in these boards, is part of the equation. However, the softness is produced by strong anti-aliasing filtering. What you see is the result of unmasking it. Halos? Yes, there's a digital-look-price to pay for it.
 
"it might look like it does but its just creates the illusion it does"
Huh?
Look. I'll run this through just one more time. Hope it helps
1) All digital images are aliased at birth! It's the nature of the square pixel matrix that is the image sensor and the displays.
2) Anti-aliasing filtration via a DSP algorithm can be pretty dumb , essentially blurring the whole image equally.
3) USM is highly a controllable retrieval of EDGE resolution that does NOT recover 100% of the original data, but rather, just enough in the right places to recreate the impression of recovered details.
4) As I have mentioned before, this problem of the 'Jaggies' as we used to call stair-stepped aliasing is eliminated once an array of hexagonaly shaped pixels is invented.
5) The expensive truth is that till then, we simply have to go with the highest resolution files, and fix the issues as best as we can.
 
You must log in or register to reply here.
Back
Top