Comparison D90 framegraps with photos...

Uwe Lansing

Uwe Lansing

Veteran
We looked at the photos of the D90 at the highest resolution (4288x2848, 12MP) compared with corresponding framegraps. I know that is quite unfair, but it shows very clear where the disadvantages of this cam are:




Here is a small clip as a waste product of this comparison:
http://www.vimeo.com/2463633
 
Last edited:
It is 100% obvious you didn't use Lee's or Matt's rescaler. Also I don't get the comparison between a still image photo vs video.
 
bronxjragon said:
I don't get the comparison between a still image photo vs video.
Click to expand...

That was my request: I don't own a D90 and so I asked capice if he could do me the favour and post some framegrabs together with the corresponding photos.

My intention:

I want to compress the photos both by resolution and by filesize. That will give us


a) a photo with 1280x720 pixels and a filesize of about 300 KB

b) a framegrab with 1280x720 pixels and a filesize of about 300 KB


We would compare then the photograph with the framegrab... both out of the same camera and both with the same resolution and filesize / bitrate (about 12-22 mbps).

I have done this already (with Adobe Photoshop) and will post the results here ASAP. The result is stunning: the framegrab with the exact same amount of pixels as the photograph (1280x720) and the same bitrate / filesize coming from the same chip / camera has a MUCH minor quality than the photograph.

This shows us, how ineffective the compression during the video recording mode works and what the camera would be capable of in its maximum abilty at a bitrate of 12-22 mbps.

It shows us also, that 12-22 mbps is far enough to get good videos.... it's the bad compression that has the biggest influence in the quality.
 
cici said:
That was my request: I don't own a D90 and so I asked capice if he could do me the favour and post some framegrabs together with the corresponding photos.

My intention:

I want to compress the photos both by resolution and by filesize. That will give us


a) a photo with 1280x720 pixels and a filesize of about 300 KB

b) a framegrab with 1280x720 pixels and a filesize of about 300 KB


We would compare then the photograph with the framegrab... both out of the same camera and both with the same resolution and filesize / bitrate (about 12-22 mbps).

I have done this already (with Adobe Photoshop) and will post the results here ASAP. The result is stunning: the framegrab with the exact same amount of pixels as the photograph (1280x720) and the same bitrate / filesize coming from the same chip / camera has a MUCH minor quality than the photograph.

This shows us, how ineffective the compression during the video recording mode works and what the camera would be capable of in its maximum abilty at a bitrate of 12-22 mbps.

It shows us also, that 12-22 mbps is far enough to get good videos.... it's the bad compression that has the biggest influence in the quality.
Click to expand...


Compression is fundimentally the issue here. :) If there was a way to do a straight dump off the chip (something alla Red's R3D) and bypass all the compression the camera must apply to the video in order to deal with the amount of data, you'd have an amazing looking 2K camera for less than $1000. Now _that_ would shake up digital cinema ;)
 
Details of photo / video under EQUAL compressing conditions:


1228763411.jpg

1228763450.jpg



- SAME camera (D90)
- SAME objective
- SAME chip
- SAME time
- SAME bitrate / filesize
- SAME resolution (1280x720 for entire frame)

above: video-mode
under: photo-mode

That's what I wanted to prove - or what I wanted to see:

The picture below shows what a D90 could do with its 12-22mbps video-mode.... and the picture above shows what it actually does...

I'm even not speaking of higher bitrate or resolution. We just want a HD-video that uses the full capacity of the chip AND of the compression at 12-22 mbps.

Note: in other pictures we see, that also the dynamic range decreases dramatically in video-mode...

Hello future! - I'm looking forward to seeing you! :)
 
Last edited:
What your not taking into account is the difference in processing horsepower between what is in camera and your computer. If the D90 either had 1) more processing power or 2) more buffer memory to handle an increase in per-frame processing time, you would see significantly sharper, better looking images.
 
wow, i could have sworn there was a search button here somewhere. :-)

you have to remember that the video mode only uses much fewer pixels from the sensor, and after debayer and noise reduction it's not surprising that the quality is lower. the compression is not the only problem, probably not even the biggest one. have you tried zooming in on the live view feed? it looks just as crappy before the jpeg compression.

/matt
 
Rexbron! said:
If there was a way to do a straight dump off the chip (something alla Red's R3D) and bypass all the compression the camera must apply to the video in order to deal with the amount of data, you'd have an amazing looking 2K
Click to expand...
why 2k? the full sensor width is 4k and in video mode it's 1.2k. and it's aliased so it wouldn't look as good as the red anyway, but perhaps better than now.

/matt
 
Its misleading to save a framegrab at 300KB in size and compare it to a downscaled image at 300KB in size.

First, the original frame of video footage never had 300K of data in it! I seem to remember doing the math and it was something like 86KB on average a frame in one of my video clips. Blowing up a 86K jpg to 300K obviously doesn't increase image quality.

Second, your computer has a lot more power than the Nikon D90's puny little image processing system. You can scale things in ways that are better and cleaner in photoshop.

Third, the still photo is taken from a reading of all the lines on the sensor, the D-Movie mode almost certainly skips lines as part of it's image scaling.

The D90 is a first generation product and it has all the faults one would expect a new technology that was only ever intended as a minor add-on feature for still photogs. Still, I can shoot with it and get a better looking image than I ever did with my low-end DOF adapter and HV20- and thats all you can get in this price range if you want low DOF shots... oh, and can I rate it around 800 ISO rather than 20 ISO like my HV20 with DOF and lens, nice that!
 
I kinda like what the compression is doing in the graffiti shot. It has this surreal vibe to it. But it would nice if it was an option we could choose in a menu!
I better run out and buy a D90 now before Nikon do a firmware update and take away the grittiness
 
First: I don't own a D90 - I just asked capice to do me the favour and take some pictures, following the instructions for my intentions.

Second: of course the PC has a better compression and the results would differ a lot....

Third: well.... THAT'S what I wanted to see: HOW MUCH will they differ? - So I know, what the D90 would be capable of, if its processor was smart enough and fast enough....

We have two very different pictures taken with ONE camera. The result has the same pixel amount and the same filesize, so we know that the D90 at its best could do it.... It's neither the sensor, nor the optics that's preventing the camera to do so, but it's ONLY it's processor and the type compression.

That's all I wanted to know by seeing it. No single framegrab and no video from the D90 could show me that. You can now also try to import a still from the D90 into a timeline.... the video will look certainly much better than a video from the D90 from the same scene... I already thought that... now I know it for sure and I even know exactly to what extent.
 
cici said:
It's neither the sensor, nor the optics that's preventing the camera to do so, but it's ONLY it's processor and the type compression.
Click to expand...
i beg to differ. the sensor only permits 1280x720 (or thereabouts) to be read out in continuous electronic shutter mode, the root of the problem. the "processor" does *not* get the same data to work with in these two cases, that's a misunderstanding.

/matt
 
You must log in or register to reply here.
Back
Top