THX 1138 - Japanese letterboxed LD preservation (Released)

The 1995 Japanese LD seems to be the same transfer as the US one that was the basis of Angrysun’s earlier DVD. My aim was to produce a cleaner looking capture with proper color. I also made two alterations: I replaced the Warner logo with the one seen on the Blu-ray, and changed the the coloring of the opening credits to green.

The capture shows noticable ringing.

Video source … 1995 Japanese CLV LD, NJWL-11162
format … 704x384 anamorphic NTSC, 23.976fps, x264 @ 3168kbps
Audio source … Angrysun DVD
format … 1.0, 48kHz, FLAC
LD player … Pioneer CLD-2950, via s-video
TV tuner … Hauppauge WinTV HVR-1250
AviSynth … 5-capture median, manual IVTC, rainbow & dot crawl
reduction, depan, partial spot removal, noise
reduction, upscaling, additional chroma denoising
on selected shots
AfterEffects … color tweaks, manual dirt removal on opening credits,
Kinney logo

Final result is a two-part MKV file. PM me for links.

sample

Thanks for the work & posting! Nice sample clip!
(My old-version MPC-HC player didn't like the FLAC audio. VLC played it fine, though.)

I looked around the other THX 1138 thread to find the multiple laserdisc-versions compare caps (US, UK, JP):

THX 1138 preservations (Italian Cut available, see 1st post)
http://originaltrilogy.com/forum/topic.cfm/THX-1138-preservations-Italian-Cut-available-see-1st-post/topic/11741/

Most of those pictures have disappeared as a result of Imageshack's recent site revamp.

I did find a page that still had some compares ..
http://originaltrilogy.com/forum/topic.cfm/THX-1138-preservations-Italian-Cut-available-see-1st-post/topic/11741/page/31/
.. but those were showing crop-differences (US & JP as a single capture as they were cropped that same).

As I remember, the JP brightness/contrast differed slightly from the other two.

From the sample, I noticed that the image looked a little soft-contrasted. When I checked, it showed a used range of 16-235 rather than the expected 0-255.

Shouldn't the capture and/or the final file have been the full range? I know the old analog broadcasting required 16-235 for safe NTSC transimission (although JP standards allowed 0-235 for their version of NTSC), but in digital media, shouldn't it have been 0-255?

.

Also, grabbing a screenshot (no aspect ratio compensation), I noticed it looked very "thin". Normally, a raw screenshot would come out looking a little "fat".

For a flat-widescreen frame like this one, the picture area itself should've come out 720x316 (as it does on a remaining US/JP screenshot from the other THX 1138 thread). The screenshot from your sample is 704x384. When I test-resized it to 720x316, it looked proper again. Is there a resizing issue here?

Spaced Ranger said:  When I checked, it showed a used range of 16-235 rather than the expected 0-255.

Shouldn't the capture and/or the final file have been the full range?

No.  The TV range should be used.  Encoding with the full range would have been a big mistake.

The AR of the sample is indeed messed up though.

Spaced Ranger: 

The image resolution is 704x384, and has an intended pixel AR of 40:33 (anamorphic widescreen NTSC). So the playback image size should be 704x317. That's what I set as display width/height in the MKV file.

Also, I believe most players expect content to be encoded with 16-235 luma range by default, because that's broadcast standard. With both MPC-HT (Win7) and VLC (OS X) the colors looked fine to me. Also both player resized the video correctly.

May I ask what player do you use?

In this system, I mainly use Media Player Classic - Home Cinema, but it's an earlier version that runs on this older OS. On the few occasions when MPC-HC fails, VideoLAN's VLC media player (also an earlier version) does the job.

No need to worry ... both players displayed the video properly sized. But I was surprised to see MPC-HC's "save image" of the file's original elongated form, sans borders (standard DVD anamorphic is 720x480, with the borders). I thought you may have made needless resizings. However, I did miss the mention of "upscaling" in the description, which mislead me in what I thought you did. Technically, then, this is not a preservation. Rather, it's a restoration from a laserdisc. That was the confusal factor. Sorry.  :)

And, yeah, that whole IRE thing tends to be more theoretical with me not doing much in the way of mastering. So I took a quick refresher course:

"IRE settings"
http://forum.videohelp.com/threads/201765-IRE-settings

So how does NTSC-Japan's IRE 0-100 (0-235) fit into this laserdisc capture for NTSC-US IRE 7.5-100 (16-235)? In the capture or post-processing?

So how does NTSC-Japan's IRE 0-100 (0-235) fit into this laserdisc capture for NTSC-US IRE 7.5-100 (16-235)? In the capture or post-processing?

 That's a good question. The capture levels I used were set using the color bars and ramp from the Video Essentials LD. I think AfterEffects converted the image to RGB using 16 as the black level, and it looked fine. Actually I lowered the shadow blacks a bit because blacks were almost always above 0%.

One possible explanation might be that the transfer was made in the US and they reused it in Japan without conforming it to IRE 0 black.

IRE should stand for It Ruins Everything, because even the "pros" don't know what to do with it:

at http://forum.videohelp.com/threads/201765-IRE-settings/page3
lordsmurf said:

I was recording a show this weekend, and the IRE of the cable broadcast changed THREE TIMES within the hour long block of programming. About every 2nd commercial break, it would return with a new value. These were VERY VISIBLE changes too, not slight ones you'd need a meter for.

.

Anyway, about those picture ratios ...

I was trying an HSL test (laserdisc's color Hue & Saturation with DVD's resolution Lightness) when I noticed that something was missed, which AntcuFaalb had come across with his US laserdisc captures in the other thread:

http://originaltrilogy.com/forum/topic.cfm/THX-1138-preservations-Italian-Cut-available-see-1st-post/post/718431/#TopicPost718431

The short story is ... another Lucasfilm problem. Their DVD release picture-area image had been stretched by 8 scanlines -- as confirmed by a 16mm print and the US laserdisc!

                                   DVD anamorphic screen-cap -- with 8-scanline reduction overlapped

The original picture area is 720x368 after trimming away the 2 gone-grey top & bottom scanlines. But the image itself had been vertically stretched by 8 scanlines. It is actually (photographically) 720x360, as shown by the overlap.

How does this size up with the JP laserdisc project? The eyes are used as natural targets for aligning the DVD & LD frames. (I left the DVD reduction area blank and colored the DVD black-bars for a more-obvious overlap demonstration.)

               DVD anamorphic screen-cap resize -- with LD anamorphic up-rez transparently overlapped

Aside from the JP LD's different cropping, the project's image is too tall at 704x362 (not including the top & bottom black-bar segments with the 2 gone-grey scanlines).

At this finished point of the project, the actual picture image must be reduced to 704x349 to match. (It probably will be slightly different during project reconstruction. Also, for this type of release, there is no need for black-bars and they probably shouldn't be included at all.)

                    DVD anamorphic screen-cap resize -- with LD anamorphic up-rez resize overlapped

csd79 said:

AviSynth ... 5-capture median, manual IVTC, rainbow & dot crawl reduction, depan, partial spot removal, noise reduction, upscaling, additional chroma denoising on selected shots AfterEffects ... color tweaks, manual dirt removal on opening credits, Kinney logo

BTW, I was wondering if you'd go into more detail on the AviSynth plug-ins used -- which ones or combinations w/ settings for each fix-up type ... that sort of thing.

There doesn't seem to be any independent sharpen. Because of the de-noising and upscaling, I would think sharpening would restore any crispness lost during those fix-ups, as well as from the LD source. My wondering was because I made that HSL reconstruction mentioned earlier (which injects DVD resolution) and found that mild sharpening brought the laserdisc image noticeably closer to the DVD. Of course, I'm using a dubious-quality paint program function. But there are AviSynth sharpeners that work without introducing such new, obvious artifacts. LimitedSharpen() is a well regarded one.

                    DVD (anamorphic expanded)                                  LD (H-S) + DVD (L) = H-S-L recombination

             LD restoration (anamorphic expanded)                                            LD w/ sharpening

Here is the archive of my work files: link.

The files 'thx-a.avs' and 'thx-b.avs' contain the averaging of 5 sources, IVTC, rainbow reduction and global stabilisation (by G-force, actually). 'thx-stage2.avs' contains the rest.

My only real regret is that I couldn't figure out a good way to reduce ringing.

Regarding the sharpening, I don't think it looks good at all. But I did the MKV with a high bitrate so anyone who wants to alter it or convert it to DVD or whatever can have a good starting point. So if you want to sharpen it... :)

Thanks for your AviSynth procedure file!

I had looked around for a plug-in to handle "ringing" or, more accurately, "edge-enhancement haloing" (that's sharpening done the wrong way). The only one I could find was DeHalo_alpha() and it required allot of experimenting to get a feel for what it does when it's variables are adjusted. Using it for chromatic aberration correction, I came up with this ..
http://originaltrilogy.com/forum/topic.cfm/THX-1138-preservations-Italian-Cut-available-see-1st-post/post/625888/#TopicPost625888
When applied to haloing caused by edge-enhancement, I made a demonstration (the post just after the above one) on a very obvious example. It wasn't eliminated, but it was reduced. Certainly worth consideration as very little bit helps.

Yeah, the paint program sharpening is obviously poor in some areas. I definitely wasn't suggesting using that.  :D  Here's LimitedSharpen() applied to that same screenshot, before & after ..

.. with this setting -- LimitedSharpen( strength=250, overshoot=0, exborder=4 ).

Wow. "Monster" is right. It's ring-smear only to the right! (A right-ringer?)  :O  I'll definitely look into it with DeHalo_alpha().

This plug-in is more like an equation, where one plays with numbers, and less like an intuitive user interface doing specific adjustments. I played with the values some and was able to reduce the faint ringing. The stronger ringing apparently is considered as picture and nothing much happens there. But some improvement is better than none ... except that there is a slight softening effect. (This plug-in is in "alpha" and has been that way for years.)

Here is a before (left) & after (right). Zoom-in to examine the result of ..
DeHalo_alpha( rx=3, ry=1, darkstr=0.4, lowsens=25, highsens=75 )
** note: ry=1 performs no work on the y-axis (vertical) **

Why was the Warner logo replaced with the one from the Blu-Ray?

NeonBible said:

Why was the Warner logo replaced with the one from the Blu-Ray?

Correct me if I'm wrong, but I believe the one from the Blu-Ray is the actual original logo or the one from the 1978 Restored Cut.  (Kinney logo)