Resolution limitations of S16 at 1080p

Muttondraw

Muttondraw

Well-known member
My understanding is that if you shoot with a good quality B4 or S16 lens, with the intention of recording at 1080p, the actual maximum image resolution you can achieve will be defined primarily by the limitations of the optical anti-alias filter in front of the sensor. Which in real world terms means you would probably only be able to image approximately 670-750 line pairs instead of the theoretical maximum of 960 which the 1080p format allows (based on Graemes figure of an imaging resolution of >70% pixel count). Assuming this is correct and I haven't missed something (quite possible) then there should be a significant resolution advantage shooting S35 or 35mm, which would take advantage of the whole sensor and downrezzing to 1080p. In this case you could achieve the maximum of 960 line pairs.

Assuming that you may want to shoot S16 (because of weight, or because you already own an S16 lens) would there be an advantage using an S35 2x extender to blow the S16 image up to fill the full sensor area and then downrezzing the resultant 4K to 1080p? I appreciate that you would lose a couple of stops in the extender, so it wouldn't be a panacea, but on those occasions where you really need the extra resolution...

I am interested in others thoughts on this.

Martin
 
This is interesting. I have wondered the same thing and it has put me off going the S16mm route.

You can shoot 2K in S16mm, so you can get a little extra resolution.
 
Shooting full sensor is always going to give the best results. The good news is that when we were setting up at IBC, we saw a mode where one quarter of the 4k image was blown up full screen on the 4k projector, and it still looked stunning. When I view a 2k crop from a 4k image at 100% it still looks stunning. Scaled down from 4k is better, but still, it looks really good native!

Graeme
 
I don't know what S35 extender you're using but I suspect it would also have to be of very good quality.
 
What about shooting 2K and downrezzing to 1080p (instead of cropping out the 1080p)? Would the extra detail be maintained or just thrown away in the downrez?
 
You'd get a smidgen extra detail from downrezzing from 2k to 1080p, but probably not to be worth the effort. But really, detail is fine native rez. Beyond that you'd have to think will that detail survive broadcast etc.

Graeme
 
jbeale said:
I don't know what S35 extender you're using but I suspect it would also have to be of very good quality.
Click to expand...

Yes you are right, there is definitely the potential for "issues". I haven't got any particular S35 extender in mind, it is just interesting to consider possibilities like this. In theory the difference in resolution achievable should be quite noticeable (assuming the lenses were up to it).

I presume the process would lose 2 stops but in theory wouldn't the noise be reduced when the 4k image is downrezzed? would that effectively go some way to cancelling out the 2 stop loss? I sort of feel like it should.. but I may well be talking nonsense.

Martin
 
In the 16x9 format, 1080p is 1920 x 1080 pixels and 2k is 2048 x 1152 pixels.

So 1080p has a resolution of 1920 pixels across and 2k has 2048 pixels. That's only a 6% difference in horizontal resolution. I don't think anyone can really see a 6% difference in resolution.
 
I think when all is said and done, the choice will be to shoot 4k or 2k... then REDCINE to whatever you need. That way you always have a "film negative" for any potential future use. Why throw away resolution from the beginning?

Jim
 
I have done a little experiment just to get an idea of what comparative differences you might expect to see in a 1080p output when capturing at different resolutions. Obviously this model is flawed because it doesn't take into account a raft of real world issues, like the difference between lenses, scaling technologies, compression artifacts etc. So please bear that in mind. I haven't used any sharpening at all.

To model the effect of a 4K capture scaled to 1080p I started off with a 4K image which I downrezzed (to simulate the effect of the optical filtration before the sensor). I then added the resolution scale (showing TV lines) to gave some objective comparison between images. This was then resized in Photoshop to 1920 x 1080 (using Bicubic smoother to minimise sizing artifacts).

To model a 2K S16 capture scaled to 1080p I took the same image and resized it to 2048 pixel wide and then resized it again to 70% to simulate the resolution loss and then resized it to 1920 x 1080.

To model a 1080p B4 capture I took the same image and resized it to 1920 pixel wide, resized it again to 70% and then resized it to 1920 x 1080.

I have included a scaled down image of the full frame (which was mostly grey) to show that each comparative image is just a crop from the top left corner of that frame.

1161647222.jpg

1161647080.jpg

1161647040.jpg

1161646974.jpg


There is a difference between 1080p and 2K, but it is only really obvious with the test pattern, but capturing 4K and scaling down definitely could be noticeably sharper, look particularly at the skin texture, the upper lip and the highlight in the eye.

Martin
 
Great example!

Actually for me the most obvious change was the boy's eyelashes, but yes, you can see what a difference it makes. Nice work!
 
You must log in or register to reply here.
Back
Top