KVCD: Background Blocks in video?
 

Hi all!!

Right now I'm experimenting with Fight Club (PAL version) and squeezing it on 1 99min. CD-R. After lots of compares, I think it looks best at 704x576 (!!!) resolution and a slightly larger Luma Blur factor (0.2). ( I got a CQ of 63...)
the movie is pretty long (133 minutes) but widescreen.

Well of course there is to say: Kudos to all of us for producing such a magnificent technique to get such a good quality!!
But you all know, as quality increases also does requirement. :D
So I know this is almost perfect, but there are two things I'd like to get rid of.

1. Ghosting. I know this is due to the noisy DVD, but especially in dark scenes with sudden movements, you can see that some of the noise is enhanced and "pulled" with an object whenever it moves. I guess there's not much you can do about it. Is it? :roll:

2.Blocky Backgrounds. This is more annoying to me. I'd love to have a more stable background. This is maybe due to the kinda low average bitrate, but as I see it, with lower resolutions the blocks get larger, and I haven't found a way to get rid of them completely anyway.

So what would you suggest? Adding another filter (VagueDenoiser maybe?) or sacrifice some more sharpness by using an even higher value of Luma Blur? So any hints are welcome! :wink:

-Andy

GREAT MOVIE! One of the best ever made..

I did it with kvcdx3 and 112 kb/s, but there was quite some mosquito noise and as you mentioned, strong blackblocking.

If your DVD standalone supports NTSC you could do a PAL -> NTSC conversion (squeeze out every bit you can :D )

1) :oops: No Idea, sorry!

2) Try Telecine() (from the Decomb filter package) and PeachSmoother. I've found out that Telecine removes grain (which might be the cause of the blackblocking) and mosquito noise quite efficiently :!:

Well... thanks for your reply. You mean I should use Telecine() on an already de-interlaced source? 8O
Well, it might be worth a test. I'll try it out. :wink:
So what settings would you suggest for Peach, and where would you put it in the script?

Hi Jellygoose (and new_bee),

My $0.02: Peach has an inherent flaw with how it handles vertical motion(unless the author has recently fixed it), so generally speaking, it's really not a very highly-endorsed filter by too many people here at kvcd.net. There was a long thread in the KDVD forum awhile back called "Recommended KDVD Filters" (or something like that) in which some examples of this flaw were described and posted (these video samples may be gone now, I'm not sure). People who contacted the author of Peach didn't receive much of a response either...first he said it was probably due to user error (which it wasn't) and simply pointed to the docs...then he eventually (months later) came back and said that there was probably just too much motion for Peach to handle properly or something of that nature (again, I don't remember what he said exactly, so don't quote me on it...I really didn't pay much attention since I personally had no intention of using the filter again). Don't get me wrong, I'm not just trying to bash Mr. Dubb or anything here, and it seems like it would be a great filter if it weren't for the motion bug. But use it at your own risk if you've got a keen eye for motion details.

VagueDenoiser seems to be the current popular choice around here when some extra compression is needed (it's quite slow though from what I've read). I've never used it myself, so I can't comment firsthand on the results...sounds promising though. Do a search for it and you should find some more info. Good luck.

-d&c

Re: Fight Club and Background Blocks... Share Your Thoughts!
 

What script do you use? :?:
This is only my opinion: new scripts are great, but with some dark movie, there is nothing better than "blockbuster" filter.
With this source (probably I'm crazy), I continue to use old script, based on "Ross Thomas" filter, and the result (related to blocky) are much better. :wink:

-break

Thanks for all your replies...

Ok, I guess I'm gonna dump PeachSmother too, after 45 minutes of testing it. 8)
I first tried it with a very strong setting, and I could clearly see the motion issues you were talking about dazed, although in still scenes the filter really shines! well I lowered the value, and it got better, but compression also went down pretty fast, so I didn't see anymore use in it.
I haven't tried that Telecide() trick yet, although I doubt that it will somehow affect the clip in a good way. However I'll try it right now! :D

After that I'm gonna test Vague Denoiser! 8)

As Fight Club is progressive PAL (Zone 2 version), you won't need any telecide().

That's what I thought. But although I don't need it, someone suggested to put it anyway, as it is said to reduce some background noise too.

From Decomb Help

I can't see how grained and combed frames compare, but it surely helped with MinorityReport..

and if a frame is combed, it is deinterlaced, otherwise it is not touched.
So Telecide() will not affect my movie at all... :roll:

I don't know, maybe it sees grain as combing..

Alright I'll give it a try! :D

Long ago, dazed&confused wrote
My $0.02: Peach has an inherent flaw with how it handles vertical motion(unless the author has recently fixed it)

Hmm... Could you describe the effect? As the author, I’d like to find out about any problems with the filter.

Hi Lindsey,

You've already seen and tried to address the described problem in this old thread: http://www.kvcd.net/forum/viewtopic.php?t=3136 .

Sorry, but I had already lost interest in that old thread by the time you had had a chance to respond, so as I mentioned in this thread, I really hadn't paid very close attention to your final posting by then...like many people here, I had already moved on in search of that "magical" filter that would give us the huge compression gain with no repercussions that we were all looking for (back then, we were still using some fairly strong temporal filtering for compression's sake, at the expense of some quality...much different to today's approach of the current 'Optimal Script'...KVCDs have come a long way in the past several months :) ).

Anyways, going back and re-reading your final thoughts in that old thread, I see that I probably could have chose my words to Jellygoose and Newbee better, had I not been going from memory. I probably should have said something more along the lines of Peach (or any strong temporal filtering in general) not being well-suited for the kinds of high quality, high-motion dvd sources being converted here (at the very least, not with Peach's defaults). In any case, at the time, I did think it was probably relevant to mention that old thread to them, and that was the main intent of my response to them here. I do recall doing some fairly extensive testing with Peach's parameters way back when (long long ago, in that galaxy far far away :wink: ), including the Dot and Readout functions, but nothing seemed to totally eliminate the movement problem that was previously discussed (although I only tested it on the 1 source, so that could have been the problem as well)....so I guess you can take that for what it's worth (if anything :wink: ). Probably just boils down to using a strong temporal filter in the wrong situation I suppose. I'll try to give Peach another look if/when I ever do a noisey vhs capture, but those are true rarities for me...and, I know, call me lazy :lol: , but I'm actually already way too happy and content with the results of the current kvcd 'optimal script' to do any substantial new testing of my own.

Regards,
-d&c

dazed&confused,

I’m not going to try to convince people what filters they should use — I just want to make sure that any bugs in Peach are squished, and that people understand how to use it.

I probably should have said something more along the lines of Peach (or any strong temporal filtering in general) not being well-suited for the kinds of high quality, high-motion dvd sources being converted here (at the very least, not with Peach's defaults).

You could even put that a bit more strongly — On pure moving material, you shouldn’t use any temporal smoothing. Fast or slow motion doesn’t matter much — Actually, slow motion material tends to be a bit worse than fast.

I do recall doing some fairly extensive testing with Peach's parameters way back when (long long ago, in that galaxy far far away ), including the Dot and Readout functions, but nothing seemed to totally eliminate the movement problem that was previously discussed (although I only tested it on the 1 source, so that could have been the problem as well)....

The single source could well be the cause — Both the “Peach Smoother is messed up” threads reference clips where the picture is constantly in motion. Peach is much worse than useless on pure moving material — It needs to estimate the amount of noise in the picture to be useful, and that isn’t possible if the whole picture is always in motion. It can handle motion fine, but only if it can first get a good noise estimate.

I wouldn’t describe Peach as a strong temporal smoother — It’s acutally a very subtle smoother when it infers that the picture is low on noise.