spoRv

Great!

If 704 is related to 640, then 640/2.35=272... if you have a picture 274 pixel height, then 274x2.35=643.9

So 704/640=1.1 then 643.9x1.1=708.29 - that is (a bit) nearer to 712 than 704... (^^,) - so, at least you should crop at 708 than 704.

Also, dicrepancy is tiny, so I'd prefer to have a really small error in aspect ratio, retaining more picture, than having a "right" aspect ration, losing some details that could be precious - as much as a little bit than 1% could be considered precious, of course!

Found, sorry, forgot to mention (here)! (^^,)

What about a "proper" Moviebarcode website? So it will be possible to understand a color grading of a movie at a glance!

Jetrell, you would add "Chicago" to the list - I've already updated the corresponding DTS list on fanres.

Yes, basically this was the idea; of course, it could be used for sources other than NTSC and PAL, but I liked the name! (^^,)

That's great, thanks a lot; community will be happy to read it - me too, of course!

DrDre, I'll be glad if you would like to take some time to write an article about the SuperResolution method on fanres! You have certain more knowledge about it than me, and it will be a great thing to have an essay, written by someone who actually uses it, and not from a software house... thanks in advance!

Still don't get the connection between 704 and 640, when both 704 and 720 are valid DVD resolutions... may you explain your decision?

I leave the task to DrDre, but sometimes I like to post my own comparison, just to spur him to do his best! (^^,)

About AR: I love to have as much picture as possible, and I hope DrDre will change his mind, and use all the sides information, becase they are mean to BE there! On the height: 548px is the perfect size, as it's exactly 2x the GOUT, and gives better results and less artifacts on the vertical direction.

I was there, hanging around, some minutes to waste, so... here you are my comparison:
http://screenshotcomparison.com/comparison/125694

DrDre, Zyrother, thanks!

@DrDre: I think we coud say, about SW versions, that the older, the better!

@Zyrother: I used two old clips captured years ago, one is a PAL LD - can't remember which one, UK? German? French? - and the other is the Japanese NTSC LD; aligned spatially and temporally, then upscaled and "refined"; you should take a look at the PaNup technique - link is in a previous post, go there, signup so you can read further info.

I thought about this, but as you can see, there is not much interest about '97SE; even if I must admit it's not my preferred version, it should be preserved nonetheless, as there is no digital retail version of it... but it will take a lot of time, and patience, and HDD space, all things lacking lately... (^^,)

Well, I worked all the afternoon just to sharp a bit the image of the laserdisc - hope it's enough, and not too much:

MagicSR V1 Vs MagicSR V1.1:
http://screenshotcomparison.com/comparison/125660

waiting for your comments!

Great news!

Have you thought about the phase inversion trick to cancel out any chroma errors?

I learned an important lessons during the years: when you get angry, you should get calm, before or after, and meanwhile you waste good time that you could dedicate to something more important, so... keep calm, and pet your cat! (^^,)

(and don't forget to use some emoticons here and there...)

Thank you, "old" pal... or it was ntsc?!? (^^,)

zee944 said:

I've seen the site Laserdisc Master posted. It's old stuff.

Infognition site, well, that's old, of course... what's about the PaNup technique? Spend two minutes to register, and five to read it, and let me know what do you think.

About SuperResolution: when used to upscale a previously downscaled (by power of 2 only), there is no resizer that is able to reconstruct details like SuperResolution can do - try it for yourself using any clip from Infognition, or just downscale any video by 50%... so, it has its advantages, that, accompanied with careful chosen filter, and/or coupled with other upsize method, delivers a very good result.

On one thing we agree: "we" (the community) need to state which is the best (apart SR, of course) upscaling method, and then use it as THE comparison... but the fact that there is (and never will be) no perfect upscaler (some reveal details, while adding aliasing, another is a bit too blurry, the other has halos) will not help... so, at the end, it's not to find out which is the absolute best, but which is the best compromise, and here we are talking also by subjective POV...

Well, now let me come back to my "pefect upscale method" I'm developing... CSI, I'm coming! (^^/)

As Zyrother stated before, this is not a competition like "mine is bigger than yours" type... what did you think? I was talking about the TV! (^^,)

Again, I'm happy to find here - and starting to find on fares, too - exciting technical threads; it's great, because it stimulating the formulation of new techniques, or, like this very thread, coming back to an abandoned one, and see it under a new light! Thanks DrDre!

zee944 said:

Can you tell me what should I see? On the forum link, I can't find description about your method, nor the screencaps of The Abyss, The Thing etc. the posters talk about. Is it about image registration and merging?

Sorry, my fault... you should be a registered member to read any first posts; sign up there, is free! There are other interesting technical articles, your feedbacks are very welcome.

I've taken a look at the SW comparsion. Do you say it's as big improvement as I achieved on the Twin Dragons caps, or better/worse?

Difficult to say... I think you could open a new thread - here on OT, or on fanres - inviting people to judge and discuss about your cleaning technique; you'll receive feedbacks, and I'm quite sure this will lead to improve your result, if you are still not satisfied with it!

About color grading: it's really a preliminary version!

About bicubic: it's a simple resize filter but, given high quality as source, it produces nice results, but not comparable to better upscaler - I always use bicubic for comparison because it is supposedly used by many TV sets, in particular the cheaper ones, to upscale lower resolution to panel resolution.

DrDre said:

@ Laserdisc Master

I certainly agree combining different methods has the best chance to achieve superior results. In fact my current version of "super resolution" now uses multiple other filters to enhance the results further. One of the great things about your algorithm is that it is able to greatly enhance SD material by using custom build methodologies. I think this is very important, because I believe there's a limit to what sharpening and denoising alone can achieve. Your upscales of the LD are a good example. The enhancement is astonishing, and I'm very sure you could not have gotten that far with sharpening and denoising alone. 

 Thanks, you are right; hope to have better captures to work with soon!

OK, I resurrect this thread after years because I think I have achieved quite good results - if you think that I used a clip taken at the time, and with same hardware and software setup, of the ones used for the previous comparison...

This implement PaNup & SuperResolution, and... well, without much further ado, go here for more info - can't copy & paste again those links... (^^,)

SuperResolution works very very well with downscaled video - in particular reduced by 2 or 4 times; take a look at Infognition comparison here; it's difficult to achieve similar results without using a temporal upscaler.

But with "real" SD material, some other upscale method, combined somehow, maybe, *could* be really similar, and occasionally spot some details where SR was not able to find - your mileage may vary! (^^,)

I don't know if sharpening, antialiasing, antiringing etc. filters could retrieve details more and better than SuperResolution alone; what I did is to use them in conjuction with SR - in a long chain - and the results *may* extract some fine details, while losing others... so, to whom has not understand this (I was one of them few months ago), there is NO filter(s) in the world which could completely restore image details from a low resolution source, to be equal to an high resolution version of the same - this is an evident assumption, but I think is good to repeat here.

Stated that, there are different methods to try to restore as much details as possible, and SuperResolution is one of them; is it the best? I don't know, because I've seen on the net WONDERFUL methods - sadly, they have so highly technical detail, that's difficult, if not possible, to find a way to use them practically - I use Avisynth like DrDre, and it will be great if we could "play" with these upsizing methods, don't you think, DrDre?

Well, I could add my own experience; I "discovered" another method to extract details, and, even if it doens't work everytime - you could read in the article why - when it works, it make wonders: the PaNup™ technique - note the ™, it's my "trademarked" upscale technique, open source, but cite the font! (^^,)

So, I used them in conjunction, and the result is astonishing! Don't believe me - again, as Mulder was used to say, "trust no 1", but take your conclusions after watching the comparisons:

PAL LD - Bicubic Vs NTSC LD - Bicubic:
http://screenshotcomparison.com/comparison/125555

PAL LD - Bicubic Vs LD MagicSR:
http://screenshotcomparison.com/comparison/125557

NTSC LD - Bicubic Vs LD MagicSR:
http://screenshotcomparison.com/comparison/125558

Next step: obtaining a perfect CSI 16x zoom in with total detail reconstruction, possibly taking them from the reflection on the pupil of a flying hummingbird - we (me and DrDre) are working on it, we'll find the solution in, let's say, two or hundred years - our clones will continue to job for us! (^^/) (^^/) (^^/)

Thank you very much, DrDre; going to post something in you thread now!

@Darth Lucas: as you stated, and as I wrote the post before, multiple captures of analog video, averaged/median'ed, helps a lot to get rid of noise inherent in the format - 3/5 captures make wonders using VHS, and LD benefits are great, too... using Happycube method (take a look at his work on this thread, or, better on this one to understand better how good it is, and to know he's pure genius!) I think it's possible to achieve a final result that potentially could be better than using the DVB sources!!!

Thanks guys!

Jetrell Fo gives me an input, and I followed it...

PAL LD - Bicubic Vs NTSC LD - Bicubic:
http://screenshotcomparison.com/comparison/125555

PAL LD - Bicubic Vs LD MagicSR:
http://screenshotcomparison.com/comparison/125557

NTSC LD - Bicubic Vs LD MagicSR:
http://screenshotcomparison.com/comparison/125558

I can't believe how much MagicSR improves the image quality; even so, it is still inferior to a simple bicubic resize of the GKar (maybe the worst of the 97 DVB recordings):

LD MagicSR Vs GKar Bicubic:
http://screenshotcomparison.com/comparison/125559

I must admit I used two small clips captured few years ago; now, taking several captures from different editions (4x PAL, 2x NTSC) and averaging them to wipe out noise, using a better capture card - if only I'll be able to let it work again... - and a REALLY better NTSC laserdisc player, or, even better, using Happycube method, I think image could be improved a lot - it *could* be really *probable* that result could be even better than DVB...

@nightstalkerpoet:
take a look at the following comparison; if you see aliasing in both, it's due to the enhancing filter, but if you see only (or more) aliasing only in the regraded version, it will be a fault of the regrading filter

DVB MagicSR Vs DVB MagicSR regraded:
http://screenshotcomparison.com/comparison/125561

it's 3:30am here, time to sleee... zzz... (^^,)

New comparisons - this time I set the image at 1920x840, refined a bit the script, and applied a preliminary color regrade.

Gkar Bicubic Vs MagicSR + BasiClean + GrainPlate:

http://screenshotcomparison.com/comparison/125542

http://screenshotcomparison.com/comparison/125545

http://screenshotcomparison.com/comparison/125548

http://screenshotcomparison.com/comparison/125552

http://screenshotcomparison.com/comparison/125553