Teranex: upscale your SD image to HD?!?! at DVinfo.net

Bill Edmunds · April 29th, 2005, 12:59 PM

At NAB, a company called Teranex had a unit that will upscale an SD image to HD. Does anyone have any knowledge of this? How good is it? Is it possible that an upscaled SD to HD image will look better than an HDV image?

Kevin Shaw · April 29th, 2005, 03:08 PM

Quote:

Originally Posted by Bill Edmunds

At NAB, a company called Teranex had a unit that will upscale an SD image to HD. Does anyone have any knowledge of this? How good is it? Is it possible that an upscaled SD to HD image will look better than an HDV image?

I've heard it claimed that really good SD source upsampled to HD resolution might look better than footage from today's HDV cameras, but (a) that seems unlikely, and (b) how much trouble and expense would be required to do that well? To my eye HDV looks great when played on an HDTV, and if I wasn't satisfied with it I'd be looking into the new JVC "Pro HD" and Panasonic "P2" cameras. I don't see any point in trying to upsample SD footage to HD unless you own a TV station or major movie archive, which I suspect does not involve most of us here.

By the way, here are some numbers to ponder in considering this question:

DV = 720x480 = 345,600 pixels per frame
Sony HDV = 1440x1080 = 1,555,200 pixels per frame

We can debate the pluses and minuses of various specific cameras relative to these numbers, but that's a heck of a resolution difference to try to overcome.

Graeme Nattress · April 29th, 2005, 07:18 PM

The Terranex units produce very good results.

Remember the resolution of the HDV Z1 may be 1440x1080, but a) that's interlaced, and b) the measured horizontal resolution is about 800.

So yes, it's perfectly possible to take, say, video from a DSR570 or an SDX900 and uprez it on good hardware and make it look better than footage from the Z1. No magic - just very good hardware (both the source camera and the magic box).

I'm working on a software solution that should produce better results still, as part of this long term R&D project I have going. It's working on stills so far, and I'm working on quicktime support so that I can stick video into my algorithm and see what happens - should be exciting.

Graeme

Kevin Shaw · April 29th, 2005, 11:27 PM

Well, the lowest price I could find for a Teranex Xantus up-converter is $60K, and a used SDX900 is $24K, so for $84K I can get an SD setup which exceeds the capabilities of a $3500 HDV camcorder? I'll get my checkbook...

By the way, how are you supposed to store and deliver the footage from all that expensive equipment, especially if your goal is simple distribution to a few end users? If I take my $84K pseudo-HD footge and compress it back to Windows Media HD at 8 Mbps, will it still be distinguishable from HDV footage compressed the same way? Okay, maybe so, but for my purposes this is a silly comparison.

HDV works, and you don't need a $60K upconverter. 'Nuff said.

Bill Edmunds · April 30th, 2005, 07:21 AM

Quote:

Originally Posted by Kevin Shaw

Well, the lowest price I could find for a Teranex Xantus up-converter is $60K, and a used SDX900 is $24K, so for $84K I can get an SD setup which exceeds the capabilities of a $3500 HDV camcorder? I'll get my checkbook...

By the way, how are you supposed to store and deliver the footage from all that expensive equipment, especially if your goal is simple distribution to a few end users? If I take my $84K pseudo-HD footge and compress it back to Windows Media HD at 8 Mbps, will it still be distinguishable from HDV footage compressed the same way? Okay, maybe so, but for my purposes this is a silly comparison.

HDV works, and you don't need a $60K upconverter. 'Nuff said.

Yikes! I'm convinced...

Graeme Nattress · April 30th, 2005, 07:26 AM

Yes, it would still look better after compression - just like 35mm film still looks better than DV even though you're watching them on a rental VHS video.

An no, it's not "cost effective", but compared to the Z1, no HD camera, no matter what the quality, is cost effective because the Z1 is so darn cheap. As you want better and better quality, you go up in price exponentially it would seem.

Graeme

Kevin Shaw · May 1st, 2005, 08:39 PM

An even better way of looking at this is that if it costs $84K to make decent-looking "fake" HD from good SD footage, for that amount of money I can buy a really nice HD camera, or wait for the Panasonic P2 and buy a couple of those with a stack of the 8GB memory cards. I think we can all agree that HDV is what it is, which is a compromise solution that's good for some things and not for others. I happen to think HDV footage looks excellent on my HDTV, but if someone else prefers a more robust solution I can understand that too.

Kyle Edwards · May 2nd, 2005, 01:17 AM

why pay 60k?

SD
http://img129.echo.cx/img129/5062/car027wn.jpg

SD upscaled to HD
http://img129.echo.cx/img129/9959/car015qx.jpg

Shannon Rawls · May 2nd, 2005, 08:49 AM

Also,

I would challenge "Z1 HD footage" vs. "SDX900 SD footage upconverted to HD footage" and shown on a HD monitor or projected on a HD screen any day of the week!

Reason I say this, is because I have shot with both cameras. Actually touch & held & used them both....not guessing or using mathematics or shooting resolution charts....but real life usage, and I'm sorry, but the Z1 native HD footage on a HD monitor will hands down look better then SDX900 DV footage upconverted to HD footage any day.

Now....if you slap a Pro35 on that SDX900 (like we did), and compare that to a plain Z1 and downconvert the Z1 to DV to make a regular DVD....then the Z1 can kiss my arse, it can't compare to the SDX 900/Pro35 package. But if HIGH DEFINITION is my goal...I would choose the Z1...and if you can get a Mini35 put on that Z1....my oh my.....SDX900 doesn't stand a chance.

- ShannonRawls.com

Graeme Nattress · May 2nd, 2005, 09:00 AM

It all depends on how good the hardware (or software) is that does the upconvert. Indeed, you cannot magic resolution out of no-where, but that's exactly what the Z1 tries to do getting 1440 out of 960. Some of the upconverts Sony was showing of DVCAM to HD at NAB were terrible - very jagged edges looking really poor, and some of the upconverts from Terranex and Snell & Wilcox were very, very hard to tell from full on HD. And indeed, it's often source material dependent, so there's no easy answers.

Interesting that you mention the mini35. Do you not find that it lowers the resolution and softens image? I know if vastly reduces DOF, but that's what it's designed for, but on top of that, do not all the extra optics and ground glass lower resolution too?? Indeed, the stuff I've seen shot on a mini35 looked beautiful, because a good picture has more to it than resolution ;-)

Graeme

Kevin Shaw · May 2nd, 2005, 08:39 PM

Quote:

Originally Posted by Kyle Edwards

why pay 60k?

SD
http://img129.echo.cx/img129/5062/car027wn.jpg

SD upscaled to HD
http://img129.echo.cx/img129/9959/car015qx.jpg

Maybe that's a bad example because of all the motion, but (a) I can barely see any quality difference between the two images, (b) neither one looks nearly as good as HDV, and (c) it's not widescreen format. Next!

Kyle Edwards · May 2nd, 2005, 10:17 PM

Quote:

Originally Posted by Kevin Shaw

Maybe that's a bad example because of all the motion, but (a) I can barely see any quality difference between the two images, (b) neither one looks nearly as good as HDV, and (c) it's not widescreen format. Next!

That's the point, it'll be the same quality as your original footage. And the original sample wasn't in widescreen. NEXT!

Graeme Nattress · May 3rd, 2005, 06:44 AM

The way to do a test is to take some HD footage, or an image from some HD footage, scale it down to SD resolution, and then scale it back to HD resolution to compare with the original. At least, that's the test methodology I'm using with my R&D.

Graeme

Kevin Shaw · May 3rd, 2005, 06:59 AM

Quote:

Originally Posted by Graeme Nattress

The way to do a test is to take some HD footage, or an image from some HD footage, scale it down to SD resolution, and then scale it back to HD resolution to compare with the original. At least, that's the test methodology I'm using with my R&D.

Graeme: that's an interesting test in terms of assessing the quality of the upconverter, but how about something more practical like taking typical DV footage, upconverting it using standard editing/production software, and comparing that to HD/HDV footage played directly on an HDTV? This is the sort of comparison we're all going to make for ourselves eventually, so it would be nice to see a thorough assessment of that. Better yet, do a "double-blind" test with several non-videographers looking at sample footage without knowing which is which, and note their comments on the perceived quality of various approaches to HD.

Graeme Nattress · May 3rd, 2005, 07:32 AM

Kevin, you're spot on right there! I guess what I'm talking about is semi-objective R&D, and that's great, but as you say, the proof is in the pudding, but in my mind, that comes later.... For myself, I do the kind of tests that I described above that compares the video to a known reference.

You could also do known reference tests too, and not tell people which is which and see if they can determine, or even do a split-screen.

I woudl think DV compared to HDV would be more of a marketing stunt than real R&D, but that won't stop me doing it....

Graeme

April 29th, 2005, 12:59 PM	#1
Bill Edmunds Major Player Join Date: Aug 2003 Location: Conway, NH Posts: 574	Teranex: upscale your SD image to HD?!?! At NAB, a company called Teranex had a unit that will upscale an SD image to HD. Does anyone have any knowledge of this? How good is it? Is it possible that an upscaled SD to HD image will look better than an HDV image?

April 29th, 2005, 07:18 PM	#3
Graeme Nattress RED Problem Solver Join Date: Sep 2003 Location: Ottawa, Canada Posts: 1,365	The Terranex units produce very good results. Remember the resolution of the HDV Z1 may be 1440x1080, but a) that's interlaced, and b) the measured horizontal resolution is about 800. So yes, it's perfectly possible to take, say, video from a DSR570 or an SDX900 and uprez it on good hardware and make it look better than footage from the Z1. No magic - just very good hardware (both the source camera and the magic box). I'm working on a software solution that should produce better results still, as part of this long term R&D project I have going. It's working on stills so far, and I'm working on quicktime support so that I can stick video into my algorithm and see what happens - should be exciting. Graeme __________________ www.nattress.com - filters for FCP

April 30th, 2005, 07:26 AM	#6
Graeme Nattress RED Problem Solver Join Date: Sep 2003 Location: Ottawa, Canada Posts: 1,365	Yes, it would still look better after compression - just like 35mm film still looks better than DV even though you're watching them on a rental VHS video. An no, it's not "cost effective", but compared to the Z1, no HD camera, no matter what the quality, is cost effective because the Z1 is so darn cheap. As you want better and better quality, you go up in price exponentially it would seem. Graeme __________________ www.nattress.com - filters for FCP

May 2nd, 2005, 09:00 AM	#10
Graeme Nattress RED Problem Solver Join Date: Sep 2003 Location: Ottawa, Canada Posts: 1,365	It all depends on how good the hardware (or software) is that does the upconvert. Indeed, you cannot magic resolution out of no-where, but that's exactly what the Z1 tries to do getting 1440 out of 960. Some of the upconverts Sony was showing of DVCAM to HD at NAB were terrible - very jagged edges looking really poor, and some of the upconverts from Terranex and Snell & Wilcox were very, very hard to tell from full on HD. And indeed, it's often source material dependent, so there's no easy answers. Interesting that you mention the mini35. Do you not find that it lowers the resolution and softens image? I know if vastly reduces DOF, but that's what it's designed for, but on top of that, do not all the extra optics and ground glass lower resolution too?? Indeed, the stuff I've seen shot on a mini35 looked beautiful, because a good picture has more to it than resolution ;-) Graeme __________________ www.nattress.com - filters for FCP

May 3rd, 2005, 06:44 AM	#13
Graeme Nattress RED Problem Solver Join Date: Sep 2003 Location: Ottawa, Canada Posts: 1,365	The way to do a test is to take some HD footage, or an image from some HD footage, scale it down to SD resolution, and then scale it back to HD resolution to compare with the original. At least, that's the test methodology I'm using with my R&D. Graeme __________________ www.nattress.com - filters for FCP

April 29th, 2005, 11:27 PM	#4
Kevin Shaw Inner Circle Join Date: Sep 2004 Location: Sacramento, CA Posts: 2,488	Well, the lowest price I could find for a Teranex Xantus up-converter is $60K, and a used SDX900 is $24K, so for $84K I can get an SD setup which exceeds the capabilities of a $3500 HDV camcorder? I'll get my checkbook... By the way, how are you supposed to store and deliver the footage from all that expensive equipment, especially if your goal is simple distribution to a few end users? If I take my $84K pseudo-HD footge and compress it back to Windows Media HD at 8 Mbps, will it still be distinguishable from HDV footage compressed the same way? Okay, maybe so, but for my purposes this is a silly comparison. HDV works, and you don't need a $60K upconverter. 'Nuff said.

May 1st, 2005, 08:39 PM	#7
Kevin Shaw Inner Circle Join Date: Sep 2004 Location: Sacramento, CA Posts: 2,488	An even better way of looking at this is that if it costs $84K to make decent-looking "fake" HD from good SD footage, for that amount of money I can buy a really nice HD camera, or wait for the Panasonic P2 and buy a couple of those with a stack of the 8GB memory cards. I think we can all agree that HDV is what it is, which is a compromise solution that's good for some things and not for others. I happen to think HDV footage looks excellent on my HDTV, but if someone else prefers a more robust solution I can understand that too.

May 2nd, 2005, 01:17 AM	#8
Kyle Edwards Major Player Join Date: Nov 2004 Location: LI, NY Posts: 274	why pay 60k? SD http://img129.echo.cx/img129/5062/car027wn.jpg SD upscaled to HD http://img129.echo.cx/img129/9959/car015qx.jpg

May 2nd, 2005, 08:49 AM	#9
Shannon Rawls Major Player Join Date: Nov 2003 Location: Los Angeles, California Posts: 853	Also, I would challenge "Z1 HD footage" vs. "SDX900 SD footage upconverted to HD footage" and shown on a HD monitor or projected on a HD screen any day of the week! Reason I say this, is because I have shot with both cameras. Actually touch & held & used them both....not guessing or using mathematics or shooting resolution charts....but real life usage, and I'm sorry, but the Z1 native HD footage on a HD monitor will hands down look better then SDX900 DV footage upconverted to HD footage any day. Now....if you slap a Pro35 on that SDX900 (like we did), and compare that to a plain Z1 and downconvert the Z1 to DV to make a regular DVD....then the Z1 can kiss my arse, it can't compare to the SDX 900/Pro35 package. But if HIGH DEFINITION is my goal...I would choose the Z1...and if you can get a Mini35 put on that Z1....my oh my.....SDX900 doesn't stand a chance. - ShannonRawls.com

May 3rd, 2005, 07:32 AM	#15
Graeme Nattress RED Problem Solver Join Date: Sep 2003 Location: Ottawa, Canada Posts: 1,365	Kevin, you're spot on right there! I guess what I'm talking about is semi-objective R&D, and that's great, but as you say, the proof is in the pudding, but in my mind, that comes later.... For myself, I do the kind of tests that I described above that compares the video to a known reference. You could also do known reference tests too, and not tell people which is which and see if they can determine, or even do a split-screen. I woudl think DV compared to HDV would be more of a marketing stunt than real R&D, but that won't stop me doing it.... Graeme __________________ www.nattress.com - filters for FCP