logo Sign In

Laserman

User Group
Members
Join date
11-May-2004
Last activity
6-Sep-2007
Posts
903

Post History

Post
#119074
Topic
<strong>The Cowclops Transfers (a.k.a. the PCM audio DVD's, Row47 set) Info and Feedback Thread</strong> (Released)
Link
https://originaltrilogy.com/post/id/119074/action/topic#119074
Time
I know how MPEG2 works, I just wanted to confirm whether you thought it looked any different on this new version between the capture and the encode, as people have been asking over in my thread wether it is fair to compare capture screenshots to screenshots pulled from a finished DVD. In my opinion if done well, the difference is so minimal as to be negligible - I just wanted to make sure you agreed before I put up any other screenshots.
I'll take it as a no, I didn't see any difference. ;^)

Another question (I'll risk the bag of poo arriving in the post), are the screenshots TR47 posted recently accurate to what is on the disc (i.e. his editing to put them all together hasn't made them noticably worse or anything)

Post
#119065
Topic
.: The X0 Project Discussion Thread :. (* unfinished project *)
Link
https://originaltrilogy.com/post/id/119065/action/topic#119065
Time
It depends on the detail in your images.
If the images are on the soft side, and you are going dual layer, then it probably won't make any difference if the bitrates are high enough. Past a certain bitrate you stop getting any improvements at all.
In general, a multi pass variable bitrate is best, just don't set the minimum bitrate too low or your quality might suffer.

The problem really occurs where you have detailed images, or a long movie and are going out single layer. Constant bit rates just can't get high enough for the amount data you want to store *and still fit on the disc*, so if going constant, you get ugly problems. (Basically because you need high bitrates for some scenes to look any good at all, but you don't have the space on the disc to do that)
This is where variable bitrates can help, you save space by having a lower bitrate on the scenes that don't really need a high bitrate, and then can use a higher bitrate for the problem scenes, and still fit it onto your disc. Have too low a bitrate on the easy scenes though and they will get ugly, so it is a balancing act.

At work we run an 8 pass variable bitrate, and then look at any potential problem areas and manually change those.
Post
#119063
Topic
***The &quot;official&quot; Screenshots feedback thread ***
Link
https://originaltrilogy.com/post/id/119063/action/topic#119063
Time
Yeah, screenshots and video are two different horses that is for sure.
Make sure you at least do a rough calibration on your TV sets and computer screens, there is free software around to at least get it setup vaguely right.

Moth3r is right though, a few of the transfers in the past have had crushed blacks and whites, which makes an image look 'punchier' but blows out the detail in the shadow areas and bright areas - no matter what you view them on.
Also the softer and more processed the image, the worse it will look on larger screens.

When looking at the stills try to look at the tonality range in the images, is there definition and detail in the bright white areas, is there detail in the dark shadowy areas?
Then look at small detailed parts of the images, can you see the intricate detail where it should be, or has it been 'smoothed over' and so on.
At first glance a crushed and smoothed picture may look "better", but on closer inspection you may realise that it is too soft and lacking in detail.

I haven't looked at the farsight set in any detail, but I did notice the crushed whites when I watched ANH on my TV.

Post
#118951
Topic
.: The X0 Project Discussion Thread :. (* unfinished project *)
Link
https://originaltrilogy.com/post/id/118951/action/topic#118951
Time
It is a good point that they aren't *exactly* like for like comparisons, but there is no intention of 'cheating' or anything going on

I haven't finalised the compression rates that we will be using yet, but I have converted my images down to the same colour space, and they are jpeg-ed so what you are seeing isn't exactly the same as the raw images, but very, very close to what will be on the DVD, and in fact the displayed images are lower quality than our hidef conversion.

On the testing I have been doing, the differences are tiny between the encoded images and the 'raw' images, and at this stage it is too much work to encode them into a DVD format and then re-extract the image to do compares. We will throw up that sort of comparison when we are in the encoding phase of the project.

The comparisons are mainly in the area of colour balance, lack of crosstalk induced video noise, and dynamic range, none of which are going to be changed by the encoding process in any real way.
The large differences shown in those compares isn't going to change much just due to an encode (and it will be nothing less than a dual layer encode)

The fine detail may change a little, but in our testing not by much at all.

The images shown so far are far from finished anyway, there is a lot more work ot be done to them, and they will only improve from here. I'd bet my left nut that the images we are showing now will be *worse* than what is on the final discs!

But yeah, it is worth pointing out for anyone who hadn't thought about that part of it.
Post
#118536
Topic
<strong>The &quot;Farsight&quot; Trilogy DVD Info and Feedback Thread</strong> (Released)
Link
https://originaltrilogy.com/post/id/118536/action/topic#118536
Time
Actually I found the IVTC built into the later versions of virtualdub never has a problem. I think it is nice and dumb and just removes based on a set pattern and doesn't try to 'guess' like the various 'automatic' avisynth filters do.
It only gets out of whack when there is an error in the source footage, in which case you just start it again from the problem frame.
Post
#118535
Topic
<strong>The &quot;Farsight&quot; Trilogy DVD Info and Feedback Thread</strong> (Released)
Link
https://originaltrilogy.com/post/id/118535/action/topic#118535
Time
Yeah, sorry Doctor M, but any project worth its salt does an IVTC process first to completely reconstruct the original 24fps film frames.
So farsight did this because he *did* know what he was doing

When the guides talk about 'The source' being interlaced, they mean the original source. In our case the original source is film, whic is 24 progressive frames per second, which was then transferred to laserdisc and made into a 30fps interlaced mishmash.
The IVTC process recreates the original film frames with no quality loss by extracting the matching interlaced halves and putting them back together.

If the 'original source' was video (i.e. it was shot using an interlaced video camera in the first place) then you cannot do an IVTC, and any form of de-interlacing does cause a quality degredation as there are no original progressive frames to get back to.

So for stuff that originated on film, doing an IVTC is a must, for stuff shot on video, you can only get a compromised image by deinterlacing and averaging and so forth.
Post
#117989
Topic
.: The X0 Project Discussion Thread :. (* unfinished project *)
Link
https://originaltrilogy.com/post/id/117989/action/topic#117989
Time
Damn these lack of avatars,

Moth3r, there is an astounding amount of video noise in that PAL image, but it does show how much promise a PAL capture will have if we can get a clean go at it - and we *are* working on it.
To get rid of the noise in the starfield, you can pretty much do it just by setting your black point to a portion of the video noise. Here is your same PAL shot from your capture with a different black point set. Original image followed by image with a different black point.

http://www.mudgee.net/ot/moth-008-crs.jpg

http://www.mudgee.net/ot/moth3r.jpg
Now it is set a little too high, but you could go somewhere in between and wipe out 80% of the noise and still get nearly all of the stars.


OK guys, as Zion said, there is another screenshot up on the x0project.com website, a little bit of Leia this time.
Remember to click through and take a look at the larger 100% sized images, as the one on the page is reduced in size.


Post
#117713
Topic
.: The X0 Project Discussion Thread :. (* unfinished project *)
Link
https://originaltrilogy.com/post/id/117713/action/topic#117713
Time
Also, what program are you using to capture with.
Lossless codecs should not look any different if using the same colourspaces, unless of course the codec or capture system has a bug, or is using a LUT during capture and mapping the pixels differently.
I'd be interested in a side by side compariosn, noting what program and card you used for the captures.

Also, if you could convert a logarith video to huffyuv (once again keeping the colourspaces indentical) and see if the degradation still occurs.

I'd be very surprised if they are different unless one does some preprocessing that the other doesn't, but it would be good to know.

The reason we chose huffyuv is it is
a) lossless
b) Stable
c) Fast
d) Capable of RGB and YUV/YIQ colourspaces

It is important when capturing such massive files that you don't drop a single frame.