Rad (1986) (Released) — Page 8

See below

I just wanted to let everyone know that Im going to be revisiting this

My plans for right now are to do a multiple pass VBR CCE DVD5 with menus imported from the original release

And Im also considering doing a 720P BD5 of this, no menu just straight movie for someone

But the picture quality should be better on the DVD since the deinterlace method on the first DVD wasnt the correct one, although it did get the job done, it did diminish the quality on end result, it also didnt put a crop on the video to get rid of the soundtrack noise on the edges, which will be corrected...Ill post some comparisons when I get the ball rolling on it

CompMovieGuy said:

... the deinterlace method on the first DVD wasnt the correct one...

You should not need to carry out any deinterlacing. Can you not apply an IVTC to the stream to reconstruct the original frames?

Thats what I was implying, I just didnt want to insult the original creator of the initial DVD

Basically what I believe is that the original DVD was made with HDTV2DVD, which imposes a Field Blend Deinterlace to everything since its 1 click and cannot tell the difference in sources being imported, basically the program is covering its own ass while giving an "acceptable" level of quality

Proper IVTC will be done along with a few other AVIsynth plug-ins

Im also doing comparisons with CCE, Procoder, and TMPG....Im just doing 2 pas VBR since thats all you can do in Procoder and I think TMPG as well. Ill post some screens and see which you guys think produces the best PQ. My guess is CCE will be the best choice given that TMPG takes a whole lot longer and the gain is minimal, ive already compared a 2 pass CCE to the original DVD, and its no comparision, significant gain in quality, obviously if I choose CCE, I will up the passes

So heres some screencaps comparing the encoders at work on Rad. In order to get exact frame matches, I used VirtualDubMod which outputed 720x480 size and i put them in 853x480 here

So heres the break down....the first picture in the sequence is CCE SP 2.7 2 pass VBR, the second picture is Procoder 3 2 pass VBR Mastering Quality, the third picture is TMPGEnc (newest version) 2 pass VBR with settings for quality, and the fourth is the source...no need to compare the first DVD since this will be better with any choice

I noticed CCE and Procoder came out nice but for some reason, it looks like TMPG didnt take my HD color correction 709->601 plugin, which is weird since I used the same script for all three encoders, you can tell in the Rad logo

Anyway tell me what you think, kinda leaning towards CCE with more passes

There isn't much to choose between Procoder and CCE. There is perhaps marginally more detail in the CCE encode (zooming in on the trees in the background of shot 3) but this may well be due to a difference in 'I' frame placement; individual screenshots are not an ideal way to rank encoders.

Also, I think I'm seeing some chroma upsampling error in the Rad logo, and the rightmost red flag in shot 3. This may be something to do with using VirtualDubMod for taking your screenshots - could you try DGIndex instead?

Well  in VDM, I scanned thru and picked all B frames for each of the screens. It was just importing the stream in, it was already created when I put it into VDM to take the screens, so no chroma upsampling should be happening. Ill take some screen tho in DGindex and put them up sometime today. I notice the two things you were talking about tho right away with the flag and the logo, for some reason it seems like TMPG just disregarded my HD color correction. I just used colormatrix.dll with these scripts in avisynth

ColorMatrix(mode="Rec.709->Rec.601")

ConvertToYUY2()

Maybe ill put up some sample clips with megaupload so that people can look at them and get a better feel, but right now Its looking like CCE with increased passes might be the way to go?

CompMovieGuy said:

Well  in VDM, I scanned thru and picked all B frames for each of the screens.

Good.

It was just importing the stream in, it was already created when I put it into VDM to take the screens, so no chroma upsampling should be happening.

The MPEG-2 stream stores video as YV12 (4:2:0), so chroma should be upsampled for display. It looks to me like VDM is upsampling assuming the frame is interlaced rather than progressive - I don't know whether this is due to a fault in the decoder or that your encodes weren't flagged as 'progressive'. 

More on this here.

... for some reason it seems like TMPG just disregarded my HD color correction. I just used colormatrix.dll with these scripts in avisynth

ColorMatrix(mode="Rec.709->Rec.601")

ConvertToYUY2()

Hmmm... doesn't TMPGEnc need RGB input? Have you tried the script with ConvertToRGB32() instead?

Well I looked at a gspot of all the clips and I noticed that the TMPG one has the S170 video icon lit, which does something to the video RBG-> YCbCr, even though nothing like this was picked in the options, and like I said I used the same avisynth script for all three encoders. Anyway heres some samples, taken from the beginging for fast paced scenes and colors, theyre about 85-90 sec samples, no audio sry, didnt see the need

I also looked this up on doom
Color Coefficients and equivalencies:
http://avisynth.org.ru/docs/english/...olormatrix.htm (Colormatrix 1.9 manual) says:
BT.601 = BT.470-2 = SMPTE 170M (BT.601 is an update of BT.470-2)
BT.709
FCC (similar to BT.601, can be handled on its own by Colormatrix v2.3)
SMPTE 240M (similar to BT.709, can be handled on its own by Colormatrix v2.3)

Is it the case that 170M is the same as 601/470, the screens are completely different, i think somewhere the colormetry is getting screwed, I can post my script if necessary

CCE 2 pass VBR sample: http://www.megaupload.com/?d=30C70NN9
Procoder 2 pass VBR sample: http://www.megaupload.com/?d=89DMYVDF
TMPG 2 pass VBR sample: http://www.megaupload.com/?d=5L6NB897

I checked the CCE sample in DGIndex and VirtualDubMod.

It does appear that VDM upsamples the chroma incorrectly (I saw the same artefacts as on your screenshots above). Not only were the DGIndex screenshots fine with respect to the upsampling, they also showed slightly different colours!

To summarise: do not use VirtualDubMod for screenshots, it's outdated and buggy.

Good to know Moth! You know you try to do something right by getting accurate frame shots and then you (by you I mean I) screw it up because the app does something to the quality you didnt expect/know about.

Well at least now I know and wont make the same mistake again.

The DVD is still in its scripting process

Currently I am collaborating with Moth3r about the best possible script plugins and about quality control

We have hit alittle bump in the road with the script but we are getting this straightened out

The new DVD will have a new TriStar opening/trailer (researching whether or not audio will be included, anyone have the VHS PM me about this) without any logos
The new DVD will have a delogo applied to it for the majority of the movie, this delogo (which Moth3r showed me) gets rid of the clear logo and doesnt result in a big blur spot (but the part with the blue logo will remain, as this couldnt be taken out without massive blurring)
examples can be seen here
http://www.geocities.com/mrtibsvideo/Logotools.html

And Im still thinking about menus if I want to reuse the old ones or if I want to totally start from scratch, but if I decide not to use any new material for the DVD, I will probably end up reusing the old menus

Hopefully encoding will start this weekend when we figure out this script problem, and everything you have seen from now is only going to get better and far from the final product

Scripting problems and file association problems (certain .dlls) have all been fixed

This project is now in the encoding phase

Currently doing a 2pass VBR CCE, then doing an eye check make sure everything is good to go and tweak what needs to be tweaked (crop, color, bitrate, and will check the delogo and will post screens)

Approx. end results:

9-10 pass CCE (extreme I know, but if Im gonna do 6, might as well wait even more, whatever), 6250 vid bitrate, only 20 secs of logo in the middle for the movie (blue logo part), new tristar trailer, proper crop (no more sound noise on the edges) obviously the 5.1 sound mix will be used

This will definitely be the best version available (maybe even possible?) with the INHD2 capture

According to Doom9 you have diminishing returns with CCE. 3-5 is recommended.

More than 5 is pretty much a waste of time.

Btw, thanks for the link for the LogoTools.  I've been using XLogo, which is dated and a bit hamfisted, but I THOUGHT about the only one around.

Doctor M said:

According to Doom9 you have diminishing returns with CCE. 3-5 is recommended.

More than 5 is pretty much a waste of time.

I dont see how more passes could diminish your results? That just doesnt logically make sense
Now I can see a minimal/insignificant gain in quality with more passes passed say 6, but thats a choice of the encoder....
I figure Ill probably still waste my time with 9 or around there

Heh, my encodes are just 2 pass in Avidemux using the "mpeg2enc" codec in the "tmpgenc" profile.