Star Wars GOUT in HD using super resolution algorithm (* unfinished project *) — Page 39

red5-626 said:

Hay there. I have a question.

Will the methods that you are using on the GOUT

work on other SD Materials?

Like old TV shows shot on SD video?

Or old 90s cartoons that are only mastered in SD?

Or on old Doctor who kinescop scans?

Somehow I doubt something shot on SD video, or a kinescope scan of any kind (remember that kinoscopes are basically a film camera pointed at a specialized tv monitor, at best having whatever the native resolution of the tv is) will have much improvement via super resolution. As for old 90s cartoons that are only mastered in SD, it depends on whether the editing pipeline had film involved or if it went straight from drawings to tape. And as for Doctor Who kinescopes, note that Doctor Who, as well as most BBC shows from that time period, used a mixture of tape for studio scenes and film for outdoor scenes. Considering that All of the Hartnell and the earlier half of the Troughton serials were recorded in LD resolution, even if they used filmed material, I doubt you'd be able to get detail from a LD source. As for the later half of Troughton as well as the Pertwee serials where the best source is kinescope, you might be able to use super resolution on the filmed scenes, assuming they don't exist separately (most likely in the case of Pertwee).

I wanted to post this earlier, but the outage happened. At least I got it off my chest xD.

And this is what I was wanting to post...

That's from my render or SR-v10, and there are many other examples of chroma-bleeding. Did you correct this in later versions of the script DrDre?

On the topic of telerecordings (/kinescope), I've never heard the term "low definition" applied to 405 and 625 line television. Especially not 625-line. I'm pretty sure you just call 405-line standard definition. Telerecordings reduced the 50i source to 25p by throwing away every other field, and recording it onto film. As the source however was videotape, it never existed in a higher resolution than what's available on the telerecording - it simply had the other fields as well - and you can't recover missing field information from existing fields.

I don't think you can get much out of such a source - the 16mm film doesn't reproduce the exact line-pattern of the source for a start, so you're working at a disadvantage in that you have a source that higher resolution than the material it contains, whereas SR is designed to work with a source that's lower resolution than the original material. Hope that makes sense.

Ah, well there's more than one way to do it, but one possible solution is to feed the SR part of your script just the luma, and use an alternate up-scaling method for the chroma.

RU.08 said:

On the topic of telerecordings (/kinescope), I've never heard the term "low definition" applied to 405 and 625 line television. Especially not 625-line. I'm pretty sure you just call 405-line standard definition.

Sorry. I meant to say that 625-line was SD and 405 was LD.

RU.08 said:

...you can't recover missing field information from existing fields.

Then how does VidFIRE work?

Someone here has access to Doctor Who kinescopes?

VidFire recreates the missing fields. It doesn't "recover them" it makes new ones. Once a videotape is telerecorded onto film the missing fields are gone. You can actually use VidFire to "recreate" "missing" fields from material that originated on film and was never on videotape, if you wanted to. Which isn't to say it's bad technology at all, just that it is what it is. :)

Depends on the TV brand as well. No-name brands will always look much worse than the top brands.

Yep, it's not complicated. Something like...

even=last
odd=eedi2().separatefields().selectodd().FrameDoubleFunction().selectodd()
interleave(even,odd).weave()

All you need is a frame interpolation function to double the framerate and return the "new" frames, I'm sure you could fine one through google. edit... you could probably use this one: http://compression.ru/video/frame_rate_conversion/index_en_msu.html

Damn good guitar player...