Canon's Secret to Achieving High Resolution in 1080p

[Petr Marusek]

Canon XL H1 Resolution from Shannon's chart: 800Hx650V progressive http://www.cinemahill.com/hidef/xlh1...s/IMG_0115.jpg

800Hx800V interlaced is from a reputable German magazine

How is Canon able to achieve such high resolution in progressive compared to interlaced resolution?

The chips are most likely using a pixel shift so the initial derived resolution that Canon starts with is higher than 1440x1080. The 1080p stream is derived from this higher resolution interlaced stream. The conversion to progressive seem to happen at the uncompressed level, because the uncompressed HDSDI, as far as I remember, includes 1080p output.

I think that Sony Z1 uncompressed output is interlaced only, which means that deinterlacing is most likely done in the MPEG2 encoder. That means that Sony starts with 1080i signal wheras Canon is likely to start with significantly higher resolution when deinterlacing.

Progressive LCD images on Z1 LCD may be simulation only, e.g. skipping one field.

[Alexander Nikishin]

Hey Petr, have you seen any low light footage beside Kakus of the XL H1? I wouldn't mind pickin up an H1 to go along with my HVX if the difference is noticeable enough. =)

[Petr Marusek]

Originally Posted by Alexander Nikishin

Hey Petr, have you seen any low light footage beside Kakus of the XL H1? I wouldn't mind pickin up an H1 to go along with my HVX if the difference is noticeable enough. =)

I haven't seen any. HVX may be OK for 720/60p, but most stuff is 1080i, so the Canon makes sense ... plus of course the Canon's sharper 1080p.

[Chris Hurd]

Originally Posted by Petr Marusek

The chips are most likely using a pixel shift...

No need to speculate about what is already known to be a fact. The chips most definitely are using pixel shift (which of course is always a good thing). I reported that months ago. Hope this helps,

[Shannon Rawls]

Originally Posted by Petr Marusek

Canon XL H1 Resolution from Shannon's chart: 800Hx650V progressive

THIS CHART is Wrong.

The camera is about 800 lines in 24F mode. It's amazing how that obscure chart has now become scientific data for everybody. Looking at the chart with my own eyes, is unquestionably no less then 750 verticle, but more like 800 to me, 775 to other 810 to others, 800 to me.

What me and Nick was reporting is what we 'seen first hand' not what we seen from that .jpg that I captured from the tape after I got home. We moved and repositioned the camera to make sure we were correct. We only rolled tape once, and that's what you see.

- ShannonRawls.com

[Petr Marusek]

Shannon, you've been saying this, implying that the camera will do 800Hx800V in 24F, if I understand right. Does it mean that the German magazine also talks about progressive 800Hx800V resolution?

Since you have the camera, could you possibly shoot the resolution chart in 1080i and 1080p and post some frame grabbs, please.

[Joseph H. Moore]

You're not really going to see a noticable rez difference in a still chart. It's not until motion comes into play that the resolution of an interlaced faux-progressive solution really breaks down.

[Pete Bauer]

Joseph, facts please...what data do you have that the XL H1's image DOES break down with motion? Have you seen the "wheel" grabs I posted last month? They are very limited and at best semi-scientific test, but showed that the image did NOT break down; nor did the beautiful stills taken of the cat by Per Kristian Indrehus: http://www.dvinfo.net/conf/showthread.php?t=57163.

But anyway, gentlemen, THIS thread is supposed to be about HOW Canon achieves its excellent results. There are other existing threads with ad nauseum arguments about interpretation of the very few posted rez chart results (including Shannon + Nick, and mine) as well as other linked grabs and clips that are available...

http://www.dvinfo.net/conf/showthread.php?t=57404
http://www.dvinfo.net/conf/showthread.php?t=56347
http://www.dvinfo.net/conf/showthread.php?t=56536
http://www.dvinfo.net/conf/showthread.php?t=55566

...plus others. Let's keep this discussion on topic and if the old news on rez charts needs to be rehashed, do it in the existing threads on that topic.

[Joseph H. Moore]

Pete,
Yes, I've seen your clips ... thanks for posting those, they were very helpful. (And, yes they also reveal some of the interlace capture artifacts.)

I'm not sure how to say this without sounding like an ass, but an interlaced capture system *has* to lose true resolution (detail or information) if the picture has changed between field captures.

Depending on how fast the motion, and how sophisticated the interpolation, the loss will be minimal or drastic.

Canon's implementation looks very good for a realtime solution. It's not going to break a sweat yielding great results on a still chart. On a moderate speed vertical pan, you'll end-up with about half the vertical rez. If it's fast enough, the motion blur will negate the loss, though.

[Tom Roper]

http://www.eidomedia.com/test

The H1 resolution advantage over the FX1 appears at the center of the frame, unlike the FX1 which maintains consistent corner to corner sharpness without the color fringing of the Canon.

Below are two side by side images taken directly from the above test and magnified to 300%. In each image, the Sony is on the left, and the Canon on the right.

From the magnified images, it's apparent the Sony is lots cleaner, has about the same detail, and none of the offensive red/green chromatic aberration of the Canon.

The resolution hype concerning this camera is totally out of proportion.

http://vsdrives.com/graphics/Resolut...risongrab6.bmp
http://vsdrives.com/graphics/Resolut...risongrab5.bmp

[Ash Greyson]

HUH?? The Canon looks a lot better to me... funny how people see things different



ash =o)

[Pete Bauer]

Folks, again, please put the resolution comparisons in the many existing threads about that topic, unless the comments directly contribute to a point you're making about deciphering the mysterious internal workings of the XL H1. That's the topic of this thread, and this discussion shall be limited to that.

[Peter Shindler]

Originally Posted by Pete Bauer

Canon said right up front at the announcement of the camera that they use a 1440x1080i interlaced CCD with horizontal pixel shift, and that it would put out 24fps using a process they call 24F via a proprietary chip they call DIGIC. How the chip does this was, is, and probably always will be unknown to anyone outside a subset of Canon employees.

I didn't and still don't particularly like not knowing at least in principle how this magic is worked, but I've accepted that they just ain't going to tell me and after a number of discussions such as this by guys like Steve Mullen and Barry Wan Kenobi, there is no new information and no new theory that gets us any closer to KNOWING. You can guess if you want to, but so far that's the best anyone has been able to do. And after all, if we COULD figure it out, it is still nothing but an entirely academic discussion.

==========================

I'm personally doubtful that a discussion about the DIGIC process will yield much useful info, but hey, ya never know. Make a thread and go for it! If the trail gets hot, I'll jump right on in there with you. Here, let's stop the meta-discussion and go back to defining and discussing camera output modes that Shannon started the thread about.

Took this from the 24f laid to rest. Pete, were you wondering how the digic chip take a 1440x1080 and converts it to 1920x1080? Or how the chip relates to the 24f mode?