Vision Unlimited/LA News

One question that periodically comes up recently surfaced on the “prosumer” group of the Cinematographers Mailing List (CML). That is “What is the best format to use for green-screen with a Panasonic HVX200 camera?” This is a different question from the general, “Which format is better?” question, because of the nature of this particular camera. Adam Wilt responded to the question with a comment about the difference in number of horizontal samples between the two versions of DVCProHD, and he is right; there is a bit more resolution available in the 1080 version. But there is a bit more to the answer, especially as it relates to compositing or green-screen.

Here is my answer…

If you are shooting 24P, the 720P compression algorithm is based on 60 frames and the 1080 one is based on 30 frames, even though you are only using 24. So more bits are dedicated to the compression of one frame in 1080 than are in 720. This is significant because the amount of information contained in the two is only marginally different, with the difference being a small horizontal advantage for 1080. The potential 33% in additional H resolution does not even start to compare to the 100% increase in the number of bits per frame available.

Here’s how it works…

Although the total number of pixels is higher in a true 1920 x 1080 frame, the useful resolution is about the same in 1080P and 720P in an HVX200 because the sensors are based on a 960 x 540 array. With offset sampling, Panasonic claims to gain up to 50% in effective resolution. Reality puts you at close to 1280 x 720, or close to the 720P limits, and the 1080P mode can’t give you any more out of the camera, although it is a pretty good conversion from 720P.

The situation for recording is a bit different, and as Adam pointed out, DVCProHD only allows you 960 horizontal samples in 720P but 1280 in 1080, so that’s where you might gain up to 33% in horizontal resolution, up to the theoretical camera limit. The vertical limit is about the same for both because you can’t get much more than 720 lines, even with the offset sampling.

In 720P/24 recording, DVCProHD applies the 100Mb/60 frame encode to each frame, even if you will throw away 36 of them somewhere in the workflow for 24P recording. There is no standard for a 100 Mb/24 frame mode, even though you are only using 40% of the frames. That is why there is no visual difference between a 24PN stream or a 24P/60 stream after pull-down removal.

In 1080P24/30 recording, the ratio is 100Mb/30 or twice the per-frame data rate, hence the compression advantage in the two frames with roughly the same information content (or at least, within 33%.) And in real scenes, the difference is probably much less than 33% most of the time.

Note that the argument still works for 720P60 versus 1080i30 recording. It’s the 60/30 ratio and the similar information content that matter. The vertical resolution may be reduced more for 1080i because of the interlace issue, while the horizontal might still be a bit higher, so the information content may be even closer between the two formats.

The bottom line should be milder compression artifacts on challenging material in 1080 than in 720 and, of course, the visual difference would be more noticeable on progressive originals. In compression, the character of the content determines the ultimate quality of the system output, so you might or might not agree with this analysis, depending on your material. Tricky stuff, this messing with eyeballs and brains…

cr/30