25Hz HDV for film blow up: 720p better than 1080i CF25?

[Dylan Pank]

In general we seem to have come to the consensus that 25 fps will still be the format of choice for a film blow-up, (smaller GOP at the same bit rate, simplicity of 25-24fps conversion sound issues aside).

Forgetting for the moment that no 25hz 720p camera currently exists, would it not be better than 1080i with CF25? After all, CF25/CF24/CF30 simply discards one of the the fields, meaning the effective resolution of CF25 is really 1440x540, a total of 777,600 pixels taken from the CCD. Add to the fact that the MPEG compression is being applied to a image area of 1440*1080, 1,555,200 pixels.

720p, assuming such a camera would really record from CCDs of 1280*720, (even with pixel shift or something similar), would take 921,600 pixels. 720p has a lower bit rate, but it is only having to compress that 921,600 with no redundancy.

...Or am I missing something?

[John Jay]

hmmm,

Whats wrong with 1080i/50 with adaptive deinterlace in post?

This way avoids recording low rez CF25 which you may come to regret downstream

[Dylan Pank]

<<<-- Originally posted by John Jay : hmmm,

Whats wrong with 1080i/50 with adaptive deinterlace in post?

This way avoids recording low rez CF25 which you may come to regret downstream -->>>

Since de-interlacing is going to involve discarding some data in post, is there not the possibility that that too would not be as good as 720p?

I guess the main problem with de-interlacing in post is, tests aside, you won't really know what your footage will look like until it has been treated. Given the processing power to do the job properly may be quite a while. Sometimes de-interlacing can look pretty bad.

720p is 59% the area of 1080i. What you've got to factor is how much resolution will you lose with an adaptive de-interlace. It might be that with a more mobile camera style, especially hand held, you maybe are losing nearly half of the resolution as the interlacing artefects are taken out, in which case 720p would be better, on the other hand, if your shots are mainly static then maybe 50i de-interlaced in post would be better.

Should this be the case, the ideal camera it seems, would be one where you could switch between formats from shot to shot, based on tests to ascertain which is preferable for a given camera shot.

Of course all this will become moot if someone does produce a true 25p camera that records to 50i.

[John Jay]

<<<-- Originally posted by Dylan Pank : <<<-- [i]What you've got to factor is how much resolution will you lose with an adaptive de-interlace. -->>>

With adaptive deinterlace you only 'lose' resolution when a subject moves in frame, the rest of the frame is full res.

Quite what the resolution of a moving subject is, is a very difficult question to answer

For example try this test

Load up in After Effects a 1956 res chart, keyframe it , then move it say 250 pixels diagonally and keyframe it again. Render this motion with 180 shutter angle over a time of say one second. (so that's 10 pixels per frame). Then after rendering choose say a frame in the middle and you will see everything is blurred - so what is the resolution of that?

(even if you half rezzed that frame it would still be blurred)

So if you say you lose resolution when a subject moves in adaptive deinterlacing - what measurement system are you using?

In my experience adaptive deinterlacing only falls over when there is field tearing due to very fast pans, something to be avoided.

[Dylan Pank]

<<<-- Originally posted by John Jay : With adaptive deinterlace you only 'lose' resolution when a subject moves in frame, the rest of the frame is full res.

Quite what the resolution of a moving subject is, is a very difficult question to answer -->>>

True, and as far as I have heard, the FX1 at least loses resolution istelf on fast moving images anyway due to the compression.

In terms of resolution trying to figure it out with numbers may be a dead end (and yes, I know I started it ;-] ) and rather it shoul be done with

<<<-- Load up in After Effects a 1956 res chart, keyframe it , then move it say 250 pixels diagonally and keyframe it again. Render this motion with 180 shutter angle over a time of say one second. (so that's 10 pixels per frame). Then after rendering choose say a frame in the middle and you will see everything is blurred - so what is the resolution of that?

(even if you half rezzed that frame it would still be blurred)

So if you say you lose resolution when a subject moves in adaptive deinterlacing - what measurement system are you using? -->>>

Well of course- though one could do try something like this -

Do your chart frame test but with any image where you can pan with a 1080 image inside, apply the motion blur and output two versions, one at 1080 with separate fields and then again at 720 as a progressive frame, then you your best adaptive de-interlace to the 1080i frame, and see which is sharper.

The results would change from one original image to another.

I also think the GOP MPEG compression also causes problems with interlacing, not just in excessive whip pans but in high contracts images and handheld footage (something a lot of filmmakers would be loathe to give up).

I've seen in Kaku's night-time bike ride footage some bizarre (I don't know what to call it so I will call it...) chroma bleed between fields in the cars' taillights when the camera was rather more unstable.

I have to ask, what program/plugin do you use or recommend for de-interlacing?

[Ken Hodson]

J.J.-"Whats wrong with 1080i/50 with adaptive deinterlace in post?"

You ever watch grass grow?

[Davi Dortas]

<<<-- Originally posted by Dylan Pank : I've seen in Kaku's night-time bike ride footage some bizarre (I don't know what to call it so I will call it...) chroma bleed between fields in the cars' taillights when the camera was rather more unstable. -->>>

That's the result of a 4:2:0 colorspace.

[Ken Hodson]

Thats interlace and lack of focus.

[Yiannis Kall]

i played 1080i sony videos at my 80'' projector and at my lcd 19'' screen and the interlaced lines looked horrible!!!
i will wait for the other companys to release camcordes with progressive format.
In my country Sony cost 5000$

[John Jay]

<<<-- Originally posted by Ken Hodson : You ever watch grass grow? -->>>

Ken, it grows real swell under 25k of HMI :)


Dylan,

I have used two ADI packages; Procoder2 and the Algolith suite, I own neither but out of the two the ADI is about the same great quality, although Algolith has the added extra of antialiasing which will be needed. Procoder2 has the benefit of not requiring the host After Effects and is I believe the faster of the two (typically 10% x realtime on a 2G P4 1080 res)

My purchasing plan is for Edius Pro 3 which includes Procoder