2k offspeed detail.

[astro]

I have just seen the results two independent 2k offspeed tests.I am concerned the footage doesn't look as sharp as I would have expected.

The post path and viewing conditions were identical to all the 4K material that has been taken care of by the post house previously. The 4k footage looks great but the 2k stuff lacks detail compared to other comperable formats under the same viewing conditons.

Its a very cursory osbervation but the concerns are vaild. Anecdotally a number of other people who have witnessed the tests have felt its looks less than it should.

Sure there are many variables involved but a perceived problem is a problem none the less.

I would be curious to know from those who have shot and viewed 2k footage, how they feel about its quality.

Thanks,
Andrew Stroud.

[Evin Grant]

It requires quite a bit of post sharpening becasue the OLPF is tuned for 4K. It can stil be quite sharp though. My guess is 1.5K actual res.

[astro]

Thanks Evin,

I only have a limited understanding of the digital world but surely you don't get something for nothing. What are the trade offs for sharpening up the image in post. Is this a common technique in digital post production?

Also if its a 2k sensor (area) then why do you guess that its only 1.5k res?

[Chris Kenny]

I only have a limited understanding of the digital world but surely you don't get something for nothing. What are the trade offs for sharpening up the image in post. Is this a common technique in digital post production?


It can introduce artifacts with certain types of subjects, and if you do overdo it, your footage ends up looking like cheap video. Which leads into the answer to your second question, actually. Most cameras do sharpening on-board. Red doesn't because 1) it's shooting RAW, and 2) you can get away with not sharpening at 4K (though it isn't always a bad idea to add a bit).

Sharpening is also almost universally applied to images from dSLR photo cameras, and Red captures images essentially the same way.



Also if its a 2k sensor (area) then why do you guess that its only 1.5k res?


Two reasons:

1) Red has a Bayer-pattern sensor (http://en.wikipedia.org/wiki/Bayer_filter). It doesn't capture full RGB data at every photo site, but reconstructs this data in post. With good algorithms this works much better than one might intuitively assume, but it doesn't quite give you full-resolution data.

2) Trying to capture detail that your sensor can't resolve results in horrible aliasing; moire patterns and such. To avoid this, digital sensors are fitted with OLPFs -- optical low-pass filters -- which filter out high-resolution detail. Nobody has figured out how to make a "brick wall" filter though -- one can can filter out only what the sensor would have trouble resolving, and absolutely nothing else. As a result, the OLPF makes the image a bit softer.

[astro]

Hi Chris,

Thats very informative. I would interested to see a film out or digital projection of a 2k sharpend footage to see how it compares with 2k scanned 35mm. Has anybody done this yet?

Thanks,

Andrew.

[jbeale]

I have just seen the results two independent 2k offspeed tests.I am concerned the footage doesn't look as sharp as I would have expected.

The post path and viewing conditions were identical to all the 4K material that has been taken care of by the post house previously. The 4k footage looks great but the 2k stuff lacks detail compared to other comperable formats under the same viewing conditons.

Can you specify what exactly is the post path? If they are (for example) extracting half-resolution data from the .r3d file, which does happen with some settings of Red Alert or Redcine, that's going to work a lot better with 4k, than with 2k material.

What comparable 2k formats are you referring to, specifically? I'm guessing that would be some form of 1080p HD and that is typically sharpened a lot. Red output is typically not sharpened at all, so the look is very different.

[Brenton]

I did the outputs for the tests Andrew is referring to.

Both were done using RedAlert, with sharpening off, medium detail, 2k Redlog exports (so using the 4K button to get full resolution of 2048*1156) into a centre cut 1080 project in Flame and Baselight for playback, grading and keying tests.

I believe both shots were 'eye focused' on the monitors @ 1:1 by the AC.

At the time of testing it was primarily for the fps motion of the offspeed. Once replayed alongside the 4k@2k, the softness issue became more apparent than if it had just been a 2k offpseed test.

With contrast & a touch of sharpening the image does improve but it is softer than I would have imagined it should be.

Now we have Redcine operating, I will be retesting some of these 2k offspeeds at the process point to see if improvements can be made.

Cheers,

B