Quality of cropped image

[Cees van Kempen]

Can a user of the F55 tell me how good the quality of the image is when a 4K image is cropped in post to 1080p. So not the full picture rescaled, but a crop from the original image. Is it good enough to use in broadcast productions? In the recent past a one chip camera with a chip similar to the size of the cropped image would not have been broadcast approved.

[Alister Chapman]

If you take a 1920 x 1080 centre crop then it looks very nice indeed, but as the sensor is bayer you will not have full 1920 x 1080 resolution, actual resolution will be about 20 - 30% lower, even so it's a very good looking image. To get full HD resolution you want to take a crop of about 2.3 to 2.5K and then scale that to 1920 x1080.

[Dave Sperling]

Cees,
'Good Enough' is such a subjective term...
Just remember, that you're throwing away a large portion of the information you've recorded when cropping. So a codec that is designed for full screen 4k may not be as effective if you're only using 25% of its information. Of course if you're recording uncompressed...
And of course lens sharpness likewise could be an issue (less so if doing a center crop, more so if using the edges of the 4k image), combined with the fact that any anti-aliasing that goes into the initial image recording is optimized for 4k sensor, not for your blown up image.
It also is a question of whether you are intercutting the footage with full 4k shots, since then the difference will be more noticeable.
So do a test -- but look at it on a BIG screen if possible. (I shot a film recently for a director who decided to make a bunch of mini-zooms in post from shots that were shot static - everything looked pretty good on small screens, until they stuck out like a sore thumb in the color correct on a 16 foot wide screen!) Remember, if everything is the same resolution it can cut pretty well -- but if you're bouncing around on resolutions it may raise some red flags!

[Cees van Kempen]

My intention is to shoot 4K voor Full HD projects now, in order to extend economical value of my footage. Plus the option to do some cropping in post. So currently I would use cropped 1920x1080 images combined with images scaled down from 4K to 1920x1080. The scaled down footage will be the majority. The film will be completely 1920x1080, so I won't be 'bouncing around on resolutions'.

Cropping to 2.5K for bayer reasons seems okay to me. However, Alan, I would say that for the same reason the 4K chip would mean a lower resolution for potential 4K projects? Isn't it? Wouldn't we need 5K bayer sensors to actually achieve 4K resolution? And what would that mean for the assumed longer economical shelflife of the footage?

I have the budget to invest in a F55 but want te be sure it is a wise investment, because of giving room to crop in post and having a longer economical value of my footage.

[Alister Chapman]

A 4K bayer sensor will not resolve 4K, so yes for true 4K resolution you do really want a camera with 4.5k+ pixels. However if we assume that we can get 3.5K - 3.7K from a 4K sensor the difference between true 4K and 3.5K will be barely noticeable in most cases and more important will be composition, noise, dynamic range etc. I suspect for reason such as sensitivity, dynamic range and ease of handling 4096 x 2160 bayer will be an acceptable standard for 4K for some time to come.

In my opinion with a 4K sensor for many shots we are already capturing more than most people actually see in real life.

[Tom Roper]

There is a very informative discussion between Scott Wilkinson and Joe Kane, (the HDTV Guru) at AVSFORUM. Although it is mostly presented from the perspective of the UHD user rather than film maker, it raises some very interesting questions about the format, especially as it relates to the aspect ratio difference, and naming conventions, as 2160p instead of the assumed 4k or UHD.

The fundamental request, is for film makers using 4k cameras planning to make 1080p converts, that sides be cropped to 3840 pixels before down sampling 2x so as to maintain an even multiple of the original, and to avoid the temptation of rescaling 4096 and letterboxing, that while preserving the native 4k aspect ratio, would not be an even multiple, and would thus introduce some softness.

I downloaded some 4k native C-log samples taken by Phil Bloom with the 1dc, for some hands-on grading and workflow. My impressions are favorable, the files are large but the workflow seems reasonable. But my first go at rescaling was slight, but noticeably softer than Joe Kane's and Scott Wilkinson's recommendation to crop before downscaling 2x as he predicted. A link to the discussion and YouTube is Joe Kane on 4K.

Phil Bloom's native file for download, once properly cropped for 2x downsample to 1080/p seems a little sharper and more detailed than my PMW350K, particularly toward the corners, if otherwise approximately the same at the center. I did notice some slight but noticeable banding in the skies if the h.264 C-log is graded too aggressively, probably owing to the wide latitude and compression. Nevertheless, I've come to appreciate the usability of the 350, as with other 2/3 broadcast cams, no longer in vogue for cinema but well evolved and complete.

[Cees van Kempen]

Originally Posted by Alister Chapman

A 4K bayer sensor will not resolve 4K, so yes for true 4K resolution you do really want a camera with 4.5k+ pixels.

So it's quite a pity that Sony didn't put a 5K sensor in the camera, like Red does with the Epic.

[Darren Levine]

Also keep in mind, that while it should look good, you're still magnifying both the lens and the sensor. if you're using top glass, then one is a non-issue, but sensors do have noise, however small it may be... until you start zooming in on it.

So the noise at full resolution technically will be half as apparent as the same clip but cropped to half resolution. We're already talking about very clean cameras, so in all likeliness it won't be very noticeable if at all, but if you were to say, want the ability to use full res and also pan and scan in post, there MIGHT be a noticeable grain difference. Again, if shot properly, cropping to 1080 from 4/5K should look great, but if you were in a low light scenario and really pushing the sensitivity, zooming in can start to show negativity.

[David Heath]

Originally Posted by Cees van Kempen

So it's quite a pity that Sony didn't put a 5K sensor in the camera, like Red does with the Epic.

Well, there are advantages to the idea of a 4k sensor. I agree with Alisters point that currently "if we assume that we can get 3.5K - 3.7K from a 4K sensor the difference between true 4K and 3.5K will be barely noticeable in most cases and .......", Basically, the leap from current 1080 to 3.5k is probably more than enough for most people at the moment anyway.

OK, one can always want more, and if a 4k sensor, why not 5k? Why not just go the whole way?

If the camera was always to be used in a RAW mode, that may be a fair point, but for a camera which can equally give a processed output, there's a lot to be said for the 4k sensor. In 4k mode, it deBayers directly to a 4k image (albeit with about 3.5k resolution). Use a 5k sensor, and there has to be a downconversion step as well, and downconversion in realtime takes a lot of effort to do a good job. (Which means expense, heat and power.)

Equally, the F5/55 are likely to be heavily used in 1080 mode for many jobs for the next few years - it's significant that Sony have put the XDCAM422 codec in there as well as XAVC. (Let alone RAW! :-) ) In which case, with a 4k sensor, in 3840x2160 aspect, it is very easy to derive the 1080 signal - a pixel is directly derived from a 2x2 photosite block.

If you really, really must have true 4k resolution there's always the F65...... For the market the F5/55 are aimed at, I think a 4k chip makes far more sense than 5k.