July 24th, 2004, 04:17 PM
What does it mean when footage is uprezzed?
View Full Version : Uprezzed?
July 24th, 2004, 04:17 PM
What does it mean when footage is uprezzed?
July 24th, 2004, 05:23 PM
DV is 720x480 (US) or 576 (europe) pixel.
This is good but for video on small screen.
If your target is big screen and/or conversion to film, there are chance you want a better picture.
Another situation would be the FX section who need hires picture to apply effects.
you can easily do this by saving your video in a serie of pictures (preferrably uncompressed tiff or bmp , no jpeg please) and increasing the size of each picture with special software (like photozoom, former S-Spline).
You can then reconvert to video.
Even when reconverting back to original size you can get improvement.
with DV, uprez is limited in a way that chrominance is half resolution of luminance, so uprez is less efficient.
additionaly you should make sure your video is progressive or deinterlaced, because scaling interlaced frame usually give unworkable results.
July 24th, 2004, 05:30 PM
The chroma on many digital video sources is at a lower resolution than the luma, and DV is particularly bad in this regard. But if you reconstruct the chroma beforehand, then things can work very well.
Most video software will uprez video taking into account the field based nature of video making de-interlacing unnecessary. If you go down the photoshop route, however, it doesn't understand interlacing so you'd need to de-interlace first.
July 24th, 2004, 05:44 PM
if you got name of softwares that uprez directly to video, i would be happy to know.
July 24th, 2004, 05:56 PM
After Effects, Shake, pretty much any compositing package will do it. I have a chroma reconstruction tool for FCP that you could use before-hand, but if you don't have that, use a DV codec that interpolates the chroma - I think the Avid one does that, or just do a 4 pixel horizontal blur on the chroma channels.
July 25th, 2004, 01:04 AM
Graeme, can you give detailed steps for the process. I have Shake 3 & AE 5.5. Do I use the chroma node in Shake? If so, what are the settings?
July 25th, 2004, 08:30 AM
On Highend 2d there are a number of chroma-reconstruction filters. The upsacling in Shake is pretty good without 3rd party software. I wrote a much more complex and higher quality algorithm for FCP that I hope to port to shake once I've learned the SDK: http://www.nattress.com/filmEffectsGNicerTests.htm
July 25th, 2004, 12:34 PM
This there a demo version of the filter available? I'd like to see it in action.
July 25th, 2004, 02:24 PM
If you're running FCP, you can get it in the demo of Film Effects. I'm hopefully going to code a better version for Shake as soon as I can get my head around the shake SDK....
July 27th, 2004, 09:03 AM
when you talk about uprez is this only done when tranfering dv to film or other large screen format only? or do you do it to improve the picture and then resize to its original size? will it aid my picture if its final output will be beta sp and dvd for TV broadcast? thanks
July 27th, 2004, 09:10 AM
Uprezzing will not help for Broadcast or a BetaSP dump. BetaSP is lower resolution than DV anyway, and a digital broadcast is the same pixel dimensions as DV.
If you want to improve the quality of DV, then you can upsample or reconstruct the chroma, but that's only of use if you're going to a verifyably higher format such as DigiBeta or HD. Using a simple linear chroma upsample will help for going to DVD though, as it has a higher chorma resolution in the horizontal direction.
July 27th, 2004, 01:00 PM
I read uprezzing and going back to 720x480 in After Effects can be done. Is this done by using a filter/effect or because of the codec being used?
July 27th, 2004, 01:13 PM
Uprezzing can be done in practically any software, AE included. AE's actually pretty good at it, as long as you turn on high quality for that layer. I don't see any advantage in uprezzing, just to return to the footage's native rez though.
Uprezzing, algorithm wise, is usually a trade off between sharpness / blurryness and edge ringing caused by the interpolation filters. I have a long term R&D project to uprez in an entirely new way, and it's working well, but I need to spend a lot more time coding it before I have something that's usable, never mind commercial!
July 27th, 2004, 01:21 PM
The guys who made www.marlathemovie.com say they used uprezzing and downrezzing in their process to smoothen out details, perhaps almost like bluring layers?
Would there not be some benefit increasing the resolution of the chroma channel before burning to DVD?
Giroud Francois's earlier post said you could get some improvement even if going back to the native resolution - I guess I'm just trying to see if I can do my own experiment but I don't know that After Effects will enchance the chroma channel on its own without applying a subtle pixel blur on it.
Any feedback would be highly appreciated. Thanks.
July 27th, 2004, 01:36 PM
I don't see any difference between normal noise removal processes and an upscale/downscale.
Uprezzing the chroma for DVD transfer is a different issue - I have a special plugin for that, but it's really for going to DigiBeta or HD, although it works very well for DVD I find it a touch overkill.
July 27th, 2004, 02:36 PM
So what are the actual steps to uprezz in AE? Which filter/effect is used?....etc.
July 27th, 2004, 02:39 PM
It's not a filter or an effect - just adjust the scale parameter of that layer. be sure to turn on high quality for that layer (the little jagged line that turns smooth when you click on it next to the layer name)
There are specialist software programs that will do better - check out Algolith for a plugin for, Shake, I think that is meant to be better.
July 27th, 2004, 02:45 PM
to Dennis H.
Uprez is working fine in only certain conditions.
for example, you should at least multiply the size image by a factor that has some meaning regarding what you will do.
for example just doubling each pixel and then dividing by two can be considered like a null operation.
I think a factor between 2.5 and 4 is a good start.
That way, fine details receive more pixel and sharpening filter can do better job.
Depending the math you use (bilinear, bicubic to name a few) you will get a more or less blurred picture. This can be positive when trying to smooth or antialias the image or a pain while trying to get more pixel for a bigger picture.
Anyway uprez by itself is not useful, it is only the base to start something with a specific result in mind (especially if you go back to the original resolution), so the tools you will apply after uprezzing will decide the result, no the uprez.
A nice way to test is to uprez only a picture and do the job into photoshop so you can see and zoom on the result.
July 27th, 2004, 08:08 PM
So if i was to scale my layer to let's say (2048X1538/ aspect ratio 4:3) then add a sharpen filter and then export the footage @ (720X480/ ratio 4:3) will that get the desired effect? Please forgive my limited knowledge on this issue. I find this uprezz thing interesting
July 28th, 2004, 12:15 AM
you are right, the process you describe will improve the picture BUT, you need to revert to the original size in a different codec than the one you start .
Probably you will start with DV codec that is pretty poor compression. after uprez, you will get losless compression pictures (like BMP) and the processing on it should improve them.
If you revert back to DV codec, the quality gained will be probably mostly lost by the DV compression except for special FX (like removing cable, blue screen, matte painting) where it works well.
Since "almost" is not "all" you could still be happy with the result.
It is up to you to decide that is why, trying on a picture only, can help because you can export it as still picture on your DV and compare.
It would be good to save into a better codec (HuffyHuv , Mjpeg 4:2:2?).
For sure if you need to tape the result and got nothing else than a DV recorder, there is no choice.
July 28th, 2004, 11:05 AM
Do you know of a good pro-level uprezzing and chroma enhancement filter or technique specifically for Adobe After Effects?
I'd like to uprezz and chroma enhance my footage before cutting to DVD.
August 4th, 2004, 10:05 AM
An upres is also known as an online. The only way to uprez DV source footage is if it was captured in a FCP offline or Avid Xpress Pro 15:1 res. This is called offline, the video quality is poor but storage is saved for capturing the most footage. When you upres the DV timeline you delete the offline media and recapture it at online res (for DV25 it is a 5:1 native).
Also Betacam SP is an 1:1 uncompressed analog 4:2:2 signal vs. DV which is 5:1 digital 4:1:1, the better quality is the Betacam SP! DV looks to the eye to be as good as SP because of it’s digital sharpness unfortunately artifacting is a byproduct of the lossy compression. Keying in DV is also difficult due to lack of color info. Also graphics and titles look poor in DV25.
FYI There is no such thing as an upres plugin & if you work in AE you sould be working at full res anyway.
August 4th, 2004, 10:15 AM
These guys are talking about taking your native 720x480 footage and increasing its resolution to an even higher number using software that analyzes it to provide a better "blow-up" than normal. So when the software does that, you can then shrink your footage back down to 720x480 and have a better looking picture then you began with.
I was hoping someone knew of a 3rd party company that made a plugin for Adobe After Effects to do that.
August 4th, 2004, 11:27 AM
How can this be done to enhance the quality of DV if you don't have the extra pixel information for a blow up to start with? You could film out to 35mm and telecine to Digi Beta or HD, but this is expensive and why not shoot on Digi Beta or HD to start with?
You could also up convert to HD with hardware or software like FCP HD color correct and down convert back to DV, but you will always be stepping on the compression or needing to render. DV is DV and it's not going to get any better, you have a standard def compressed signal with limited color info. That is why DV format is relatively inexpensive.
Also try this, put your video on track 1 of your NLE or AE and then put the same footage above it on Trk 2 and experiment with the opacity and transfer modes modes, this may yield an outcome you are looking for.
August 4th, 2004, 11:38 AM
You'd choose miniDV over HD or Beta for financial reasons, and increasing the picture quality would be done for artistic reasons. However, to answer your question on how it could help to uprezz - I've used software that does things like frame and line interpolation, for things like converting interlaced footage to progressive.
That software takes the every-other line and every-other frame data inside its algorithms to build a progressive frame. I would think that if someone used similar technology in a very complex and skilled way, that they could build a plug-in that (though it would likely be very processor intensive) would be able to interpolate your footage to expand its size without much if any loss of detail (within limits of course). If that could be done, theoretically there would then be more image detail to work with (what was added by the plug-in) and therefore, scaling the image back down to normal size may in fact produce a sharper looking image.
Affter Effects and photoshop do this to a degree, but not well, and I wasn't aware of anyone that did. This thread led me to believe perhaps someone does know of software that does that, which if it exists, I'd love to see :)
Of course, everything I said is conjecture and I could be completely wrong in my assumptions.
August 6th, 2004, 07:57 AM
Erick Yablon thanks for your informative posts i also want to ask if as you said that beta sp is a higher format than dv let say that i get a sp rec deck hook up in my computer if i have a dv footage thats 720*480 would it benefit mmy picture if i uprez to lets say 2048X1538 run a coplue correction filters on it and then resize it to 720*486 (thats sp res) and record it on the sp deck directly? or the results would be the same if i choose to tranfer dv to sp directly and let the extra 6x empty? thanks
August 6th, 2004, 08:24 AM
Erick said "Also Betacam SP is an 1:1 uncompressed analog 4:2:2 signal vs. DV which is 5:1 digital 4:1:1, the better quality is the Betacam SP!"
which is false. SP is not uncompressed. It is bandwidth limited to a lower resolution than DV and DigiBeta. It is not 4:2:2 - because the chroma bandwidth is also limited to about 1/4 of full resolution luma. Just because people capture it uncompressed 4:2:2 does not mean that SP is that!!!
If you're dubbing DV to SP, just ignore the 6 "missing" lines. Doing anything else will reduce quality.
I really like the concept of uprezzing DV to HD resultion, but I have great difficulty accepting that downrezzing it again is going to introduce any great benefit.
August 6th, 2004, 07:35 PM
DV vs. Betacam SP: 4:1:1 vs. 4:2:2, Artifacts and Other Controversies:
The Information above is provided for everyone to make his or her own educated decision on which format is more robust and which provides a better quality video signal for there production and post production requirements keeping in mind performance/cost ratio.
“The DV format offers a virtually unbeatable performance/cost ratio” this may be true but DV has definite limitations in post and that is a fact, from artifacts to chroma keying, other wise you guy would not be trying to, as you call it “upress” a DV signal.
Here is a question, if both formats had the same acquisition and post expense, which would you choose to use?
August 6th, 2004, 07:49 PM
What I'm working on in uprezzing DV up to HD, which is no trivial task... but it's made easier by the nature of DV in that it might have low colour sampling, but you always know where those sample are and how they're spacially arranged.
And interesting (and old) that article is, it doesn't tell you any different to my comments that SP is bandwidth limited, and has as bad chroma as DV (but fuzzy rather than sharp). If you blur the chroma from DV it looks like chroma from SP.
And if I had a camera for which I could put on the back an SP back or a DVCAM back, I'd put a DVCAM back on it every day of the week. But now we're really getting OT.
August 6th, 2004, 08:02 PM
your trying to make a skateboard into a formula 1 race car.
August 6th, 2004, 08:10 PM
Absolutely! Because most of use can't afford HD, but we can afford DV. My goal is ALWAYS to get the best from modest equipment. DV provides a very high quality, stable, digital picture. It may have artifacts, but they're able to be dealt with. The chroma sampling is NOT an issue with DV if you use clever algorithms. You just have to know how to decode the lost chroma info from the luma, which is a technique that you can't use on SP because it's fuzzy in the chroma.
August 6th, 2004, 08:17 PM
Rafael I'll let Graeme answer that one respectfully.
My opinion is DV is DV if you want something more robust you should start with acquisition. You will spend just as much time and money trying to “upress” DV to another format CODEC while compressing it twice as much by the time you’re done. If Graeme can invent a lossless plugin that cost less than a one off DV to SDI/HD upconverted dub than I’ll take notice.
August 6th, 2004, 08:25 PM
This is what I'm doing today on DV chroma reconstruction:
Combine that with the algorithm I'm working on for uprezzing the whole image (luma and chroma) which I posted an early test of, and you can see that I believe that I can make DV look an awful lot better than it deserves to. One of the nice things I've found so far about the algorithm is that it sharpens video without putting those nasty sharpness lines around edges, and will even remove some of the bad effects of existing sharpness lines. But I'm probably a year away from a fully working version....
August 6th, 2004, 08:37 PM
I'm wonder that R&D buget and market demand will offset an inexpensive solution. I can't fault you for thinking out side of the box.
August 6th, 2004, 09:09 PM
Why not invent a hardware cuicit that is inside of the DV camera that will do this "upress" converion realtime while aquiring the footage?
August 7th, 2004, 06:11 AM
The best thing about DV is that it's standard and affordable. If you uprezzed inside the camera, how would you store it??
Anyway, I'm not an electronics person, I'm software, so I do what I do best.
I'm self funding the R&D - it will get expensive in the end when I have to compare what I'm doing to "real" HD, but I also have some other R&D projects which are cheaper and should help fund the big one....
August 7th, 2004, 07:39 AM
how would you store it??
I guess you colud tweek the codec or step on the compression a little.
August 7th, 2004, 08:10 AM
Mr. Nattress after reading and looking at your page i think the Film Effects: G Nicer chroma reconstruction is awsome and is far better than doing nothing to it. is it better to apply this filter to an uprezzed picture let say a 2k or hd res? or will it yield the same results on a standar dv 720*480 image? also would it be sound to apply a sharpness filter after the chroma reconstruction? is really a shame that i only have a PC cuz i belive that your software is for the MAc i have premire pro and a matrox rtx100 system. is there a way that i could do this kind o job to my footage using a pc? have u ever try to write filters for virtual dub or avisynth? thanks again
August 7th, 2004, 09:44 AM
Why not invent a hardware cuicit that is inside of the DV camera that will do this "upress" converion realtime while aquiring the footage?
Erick, you seem to be so sure of yourself and why there's no point in trying to reconstruct the chroma detail, but here you clearly show that you have a limited insight in the way that the different video formats work.
The chromaresolution in DV (and HDCAM for that matter) is reduced in the encoding stage, before the signal goes to tape. Before the video is compressed in the camera, the resolution is there, but it's downsampled to cut bandwidth.
So there's no need to upres in-camera, since the resolution is already there.
The point is that the chroma of DV and HDCAM comes out very pixelated. The human eye is rather unsensitive to color detail, which is why they can get away with it. But for processing like color correction and chromakeying, you want to have that information. A good solution to reducing the pixellation of the chromachannel is to resample it using a good scaling algorhythm. Although you end up with an image the same size, it'll be less pixellated and smoother, withouth blurring.
A plugin that does this is the de-artifactor that comes with Magic Bullet Deinterlacer. Just a simple test with this shows that there's a HUGE benefit in reconstructing chromadetail on DV footage before processing it.
Obviously shooting at uncompressed HD would be prefered technically - but that's not the point.
Just to set things straight.
August 7th, 2004, 04:48 PM
My apologies Barend,
I was being flippant regarding the upress circuit, and just to set things straight, I’m not a video engineer but I do understand tape formats professor.
I question your logic:
“Before the video is compressed in the camera, the resolution is there, but it's downsampled to cut bandwidth.
So there's no need to upres in-camera, since the resolution is already there.”
-What resolution is already there? please be specific.
“resample it using a good scaling algorhythm”
-good idea! But why spend hours or possibly days rendering with an expensive plugin when you could do this real time in the cameras circuitry when it is encoded to tape?
Check out the Cannon XL2, you will find new circuitry built in to the camera to get a closer to film look when acquiring (no plugin needed), not to far off from my idea.
August 7th, 2004, 05:19 PM
Like I tried to explain earlier, the chroma resolution loss happens at the stage of compression to DV. This is just part of the DV format, that the chroma channel is downsampled. There's no way around that as long as you write to DV tape.
It's just the limitation of the DV format. And the HDCAM format for that matter. And although the XL2 may have slightly higher resolution chips resulting in a sharper image, the image that goes to tape still has it's chroma downsampled at the DV compression stage. This is what they mean by 4:1:1. For each 4 samples of luma information, only 1 sample of chroma is stored on the DV tape. So the only way around it is use a more expensive tape format, Digibeta or HDCAM SR - or start thinking about recording directly to disk...
Ideally you'd record 4:4:4 but that's a huge amount of data, so for practical reasons most digital formats use some kind of chroma downsampling.
Hope this helps.
August 7th, 2004, 08:07 PM
DVCPRO 50 is 4:2:2
August 8th, 2004, 03:52 AM
Would using DVCAM tape instead of miniDV inprove the image quality?
August 8th, 2004, 06:54 AM
DVCPRO 50 is 4:2:2
Yes, and so is DigiBeta, albeit with a less harsh compression ratio and thus a higher bitrate.
But I fail to see how this adds to your point. The XL2 doesn't record to DVCPro50. And the 50 means that it's got a double datarate (50mbps) compared to DV25's 25mbps - it's essentially one of the more expensive tape formats that I was referring to in my previous post.
Recording to DVCAM instead of DV doesn't help either. The codec for DV and DVCAM is exactly the same. The most obvious differences are that DVCAM has a higher tapespeed so it's less prone to dropouts due to dirt on the tape. And DVCAM has locked audio.
August 8th, 2004, 12:55 PM
If you think it's possible to uprez a DV camera to HD, then I'm going to uprez my HD camera to be IMAX. ;-)
August 8th, 2004, 02:11 PM
Uhm... did anyone say that?
August 8th, 2004, 02:50 PM
I said that - I think I can do a pretty darn good job of it once my scaling algorithm is working properly. (probably about 1 year's R&D away) I can already turn 4:1:1 DV into a pretty good facsimile of the original 4:4:4, which is the first step along the way of scaling DV to HD... OK - it will never be as good as shooting HD, but the better I can get it.....
August 23rd, 2004, 06:56 AM
I've been doing some more work on the algorithm I was talking about earlier in this thread. Please don't get too excited as I'm still along way off a usable product, but I thought you'd like to see how it's progressing:
Comments (both positive and negative) are appreciated.
April 11th, 2008, 03:26 PM
I look forward, Mr. Nattress, to your results.
I also would like clarification on the use of Shake, though just finding the reference will make me research this question at Apple. The only thing I can add besides questions is the mentioned use of Photoshop. There is an excellent 'format' used for fractal image growth called GF Print Pro, now on version 5. While not as perfect as capturing the image at the resolution you would like originally, it's quite good. Designed for print media, it works with both RGB and CMYK. The extreme downside is the fact that even with macro automation, you're talking about frame by frame.
One note: What drew me to this thread was that I have EXISTING footage of a subject who is dead. I need to incorporate it with HD footage. Between the footage being shot and now, there have been a few changes in DV reality- What we now call High Definition becoming more common than Standard Definition. Many of the posts in this thread seem to be smugly declaring that shooting HD to start with keeps one from being an idiot.
Does anyone really believe that what we now know as HD will not be supplanted by higher resolution cameras, indeed, by 3-D, tank 3-D and even cameras shooting for hologram projection stages? Do you even today want to compare the highest DV HD with kodachrome? How about 4X5 kodachrome? It makes IMAX seem small. I do doubt, though, that anyone could think of a reason to transport 4X5 film for motion pictures.
Jeez, what a strange rant. Anyway, I'm looking, as the originator of this thread was, for a good way to upres.
April 11th, 2008, 04:34 PM
Has anyone used this? I am fond of their Magic Bullet Editors and Look Suite, myself.