logo Sign In

Info: laserdisc captures - of resolution, resize and cleaning...

Author
Time
 (Edited)

It seems there is some kind of interest about the fact of capturing laserdisc contents, also if the same titles were used previously for other projects; indeed, the fact that different versions, copies, and players could be mixed to obtain the cleanest laserdisc capture ever produced until now, is really exciting!

So, here I am: I’m ready at last! The laserdisc players are here, as the SW laserdisc collection; there is a brand new 1TB HD that will be dedicated exclusively to this task, and really soon, a new internal HD will be installed with a clean o.s. - virus-free and with only the necessary programs, nothing more.

It’s time to start to capture, but, as I want to start right from the very beginning, I’m here (again) to ask your help - as usual 😉

I read and read and read this forum and others, and I learn a lot, any day, but there are still some things I’m not sure… so, without further ado, let’s start the thread!

Horizontal resolution:

At first, I thought to use the max resolution available to capture the PAL LDs, so 768x576 seemed a good idea (and the PAR is 1:1)… but, after cropping black sides, the picture resulted about 748x576, and for DVD or AVCHD (standard definition) I must use some kind of resize, and lose some quality… well, then I thought: “why can’t I capture at, for example, 740x576? After cropping 20 pixels, the result picture would be 720x576, perfect!”… the idea, in practice works, but is WRONG!

I discovered (after reading A LOT on the net) than ANY capture device captures at FIXED RESOLUTION, and then resize! So, it is futile to use any other resolution but the only one which my card uses. For my card, whose chipset is Philips SAA713x, I know this horizontal resolution is 720 (as stated on the official Philips documentation) but reading more, as it sample at 27MHz, 2x oversample, so 13,5MHz, the resolution is 702…

So, what is the REAL resolution my card captures? I think 720, as stated in the official document; I tested it, and about 8/10 pixels per side are black (due to overscan, I presume). If I crop, then the result is 704, good for DVD, not compatible with AVCHD… I could do two versions, one at 720x576 (with some black borders at the sides) for AVCHD, and the other at 704x576 (full horizontal pictures without borders at sides) for DVD. What do you think?

Resize, upscale and so on:

Well, let’s assume now the captured video is 720x576 25p (yes, now the deinterlaced problem is solved, and the video is progressive). This is the full picture with black borders top and bottom, and it’s aspect ratio is 4:3. Really good for 4:3 analog tv sets… but all the digital displays will resize the picture anyway…

So, it is REALLY necessary to resize the picture to make the aspect ratio 16:9 compatible? Many restorations have done this resize, but I think is not so fundamental, as also the 720x576 16:9 picture will be resized by the display…

Then now someone could argue that a proper done conversion from 4:3 to 16:9 is ever better than a real time resize done by the display… it could be true, but is also true that this very display will resize in any case, so there will be two upscale (first from 4:3 to 16:9, then from 720x576 to other resolutions like 1366x768 and 1920x1080)…

Then another idea come to my mind: why don’t resize the original picture 2.5x? The resulted video will be 1920x1440 4:3, cropped at 1920x1080 16:9; using a very good AviSynth scaler, and a bitrate of about 8mbps, an AVCHD will have a quite good image, THE SAME for any hd display, not depending on the display internal scaler… but I thought also: it’s better to have the original video 720x576 using 8mbps and let the display upscale, or 1920x1080 using 8mbps and not further upscaling? Or, just convert from 4:3 to 16:9? At least, in this case my old 16:9 analog tv set will display more lines than my old 4:3 analog tv set… 😉

…maybe the idea of resizing is not so good - also if my latest tests with Super Resolution of Video Enhancer (which uses spatial and temporal interpolation) is quite good, as the upscaled video looks like with some tiny particulars not so clear when seen at original size - but this is really subjective, and maybe only my imagination!

At the end, a simple conversion from 4:3 to 16:9 using the best AviSynth script will be the optimal, final solution. Or not?!? What are your opinions?

Cleaning the video:

After all the negative comments of my so-called whitemagic AviSynth script, I decided to <span style=“text-decoration: underline;”>not use it</span> to clean the captured video. But some sort of cleaning is surely necessary, as the laserdisc (which I LOVE), despite the fact that was the best consumer analog standard definition video format, it’s still analog composite video…

What it MUST be used is a script which takes comparison of different captures, and clean drop outs and other glitches like that. Maybe median could be good, but I used TOOT and the results seemed better to me. Any other good method?

Then, some denoise, and, if necessary (with the “pyramid” set for sure) color correction. Anything else? And, when applying those ones? After the (eventual) resize, or before?

The first thing - horizontal resolution - is the most urgent, as if I use the wrong one, I must capture again… and with so many hours, sides, discs, versions, LD players… I couldn’t think about that!

The others could wait, as they will be applied after the capture.

Any eventual other hints and comments are welcome. Thanks in advance to all of you!

Sadly my projects are lost due to an HDD crash… 😦 | [Fundamental Collection] thread | blog.spoRv.com | fan preservation forum: fanres.com

Author
Time
 (Edited)

I believe that the procedure is the following:

1. Capture at 720x576.
2. Crop 8 pixels from each side to account for overscan, leaving you with 704x576.
3. Use PAL's 12:11 PAR to get square pixels, so (704 * 12)/11 = 768, leaving you with 768x576.

It's preferable to crop and resize in one invocation so as to not cause sampling issues. The following would do the trick:

capture.Spline36Resize(768, 576, 8, 0, 704, 576)

Edit: Make sure to not do any more horizontal cropping after this!

Edit 2: Also, if you're interested in cropping out the black borders on the top and bottom, please let me know so that I can modify the above code.

Edit 3: Do you want me to give an example using PAL GOUT? I can include information on how to make it anamorphic widescreen as well.

A picture is worth a thousand words. Post 102 is worth more.

I’m late to the party, but I think this is the best song. Enjoy!

—Teams Jetrell Fo 1, Jetrell Fo 2, and Jetrell Fo 3

Author
Time
 (Edited)

What color-space are you capturing in? I would suggest using 4:2:2; e.g., YUY2 or UYVY.

Edit: Do you plan to distribute raw captures? It'd be worthwhile to do in case you make any post-processing mistakes.

Edit 2: Also, what codec are you using? I prefer Lagarith for raw captures.

A picture is worth a thousand words. Post 102 is worth more.

I’m late to the party, but I think this is the best song. Enjoy!

—Teams Jetrell Fo 1, Jetrell Fo 2, and Jetrell Fo 3

Author
Time

_,,,^..^,,,_ said:


What it MUST be used is a script which takes comparison of different captures, and clean drop outs and other glitches like that. Maybe median could be good, but I used TOOT and the results seemed better to me. Any other good method?

I prefer the following approach:


function MedianOfThree(clip c, clip l, clip r)
{
return Interleave(l, c, r).Clense(reduceflicker=true).SelectEvery(3, 1)
}

However, you have to make sure that the three captures are 100% lined-up spatially and temporally.

A picture is worth a thousand words. Post 102 is worth more.

I’m late to the party, but I think this is the best song. Enjoy!

—Teams Jetrell Fo 1, Jetrell Fo 2, and Jetrell Fo 3

Author
Time

AntCu, thanks for the fast answers!

As my computer is "only" an old 1,86GHz Core 2 Duo, I think Huffyuv is the best codec to use with it, but I'm open to Lagarith - I know it compress more, must do some tests, when the new primary HD will be installed.

Using Huffyuv, the color space will be YUY2, as Huffyuv's default color space is YCbCr and laserdisc's is YUV, so all of them are 4:2:2.

But, why resize at 768x576 if I plan to do DVD or AVCHD?!? I know its PAR is 1:1, but it is off standard...

By the way, crop and resize in one instance, or crop first and resize after, operating with a codec like Huffyuv, is the same, as the codec is lossless... isn't that right?

About distribution: I just finished to calculate the size taken for each trilogy, and if my calculation is right it will be about 240GB using Huffyhuv!!! With my poor ADSL it will take forever to upload the files! Also, with "only" 1TB of hard disk dedicated space, I could capture (if I'm lucky) at best four trilogies... maybe it's better to capture one episode at time - so I could store up to 12 captures - not that I need all of them, but considering versions, copies and laserdiscs, I think I could reach 9 captures easily... then, after that, I could "mix" them and store as long term only the result. And restart with the next episode!

For your median script: I'll test it after the first captures are done. But did you tried the TOOT script? Yes, I know I must align spatially and temporally the captures, I've done it before - with A LOT of work, if I remember well ;-)

...using the TOOT script, the results were good.. but I used only three captures. With more, like nine captures, I could TOOT 1,2,3 then TOOT 4,5,6 then TOOT 7,8,9 and finally TOOT the results... at the end, the picture will be as clean as the best laserdisc picture could be, dropouts free!

I used here TOOT as example, but I'm open to use another script, like yours. If someone else has good experience with multiple captures, please post your script!

Antcu, can I count on your help if (when) needed, for what concerns AviSynth and/or VirtualDub scripts and settings?

Sadly my projects are lost due to an HDD crash… 😦 | [Fundamental Collection] thread | blog.spoRv.com | fan preservation forum: fanres.com

Author
Time

I am sorry if I missed it but which laser disc are you going to capture?

looking for HDTV of the  Attack of the Clones and Revenge of the Sith.  Also HDTV of The Lord of the Rings trilogy

Author
Time

Troyig88 said:

I am sorry if I missed it but which laser disc are you going to capture?

All of them? ;-) No, I'm joking (and time/hd space is limited, too!)

You are right, I posted them into another thread:

http://originaltrilogy.com/forum/topic.cfm/Audio-and-or-video-captures-offered-anyone-interested/post/602893/#TopicPost602893

 

Sadly my projects are lost due to an HDD crash… 😦 | [Fundamental Collection] thread | blog.spoRv.com | fan preservation forum: fanres.com

Author
Time

Can we chat on IRC about this, Andrea?

A picture is worth a thousand words. Post 102 is worth more.

I’m late to the party, but I think this is the best song. Enjoy!

—Teams Jetrell Fo 1, Jetrell Fo 2, and Jetrell Fo 3

Author
Time
 (Edited)

Resizing to 768x576 is done as a way for you to get the true aspect ratio of what you're capturing. It's not the final size.

Typically, after following my instructions above, you're left with a 768x576 frame with a square PAR. You then crop out the black borders and, assuming all went well, you're left with the proper display aspect ratio (DAR). You use this DAR to make sure that your final size accurately reflects the proportions what you captured.

Example: Transforming PAL GOUT to an Anamorphic NTSC DVD:

1. Crop 8 pixels from the left and right, leaving you with 704x576.

2. Resize to 768x576 for a square PAR.

3. Crop 124 pixels from the top and 126 pixels from the bottom to get rid of the black borders. This leaves you with 768x326.

4. Notice that 768/326 = ~2.35. This is the DAR that we want to maintain.

5. NTSC DVD is 720x480, but when anamorphically stretched on a widescreen TV, the size is ~854x480 (16:9), so we need to solve for X in the following equation: 854/X = ~2.35 (our DAR). The answer is 362. (Please use a calculator for this and don't round!)

6. We can now use 720x362 as our final image size. When anamorphically stretched on a widescreen TV, this will result in 854x362 which is close to our DAR (2.359 vs. 2.355).

7. Now we add in black borders (59 on the top and bottom) to get the height to be 480.

Notice that most of these steps are used to determine the correct final size. You don't (and shouldn't) put them all separately in the script. For this particular example, all you need to do is the following:

gout.Spline36Resize(720, 362, 8, 124, 704, 326).AddBorders(0, 59, 0, 59)

This will look vertically stretched on your 4:3 TV, but will look 100% correct on your 16:9 TV.

A picture is worth a thousand words. Post 102 is worth more.

I’m late to the party, but I think this is the best song. Enjoy!

—Teams Jetrell Fo 1, Jetrell Fo 2, and Jetrell Fo 3

Author
Time
 (Edited)

(due to overscan, I presume).


It's not overscan, but nominal analogue blanking. When calculating PARs and DARs, only the central 704 pixels (702 for PAL) should be taken into account.

It's simpler to ignore width completely when trying to work these things out. To go from PAL non-anamorphic to NTSC anamorphic just multiply the height by the NTSC/PAL ratio (480/576) and then by the anamorphic ratio (4/3).

326*(480/576)*(4/3)=~362.

DE

Author
Time

Darth Editous said:

 

(due to overscan, I presume).


It's not overscan, but nominal analogue blanking. When calculating PARs and DARs, only the central 704 pixels (702 for PAL) should be taken into account.

It's simpler to ignore width completely when trying to work these things out. To go from PAL non-anamorphic to NTSC anamorphic just multiply the height by the NTSC/PAL ratio (480/576) and then by the anamorphic ratio (4/3).

326*(480/576)*(4/3)=~362.

DE

 

Yay! We got the same answer using two different methods.

A picture is worth a thousand words. Post 102 is worth more.

I’m late to the party, but I think this is the best song. Enjoy!

—Teams Jetrell Fo 1, Jetrell Fo 2, and Jetrell Fo 3

Author
Time

If you have several captures you could try interpolating the frames with MflowFPS to remove dropouts etc.  Since each frame is identical except for the fact that it is from another capture it should only average out noise.

Luke threw twice…maybe.

Author
Time
 (Edited)

I hope my cropping&resizing information helped, Andrea!

Darth Editous has confirmed that my calculations were correct using a different method, so I hope you'll trust what I've posted.

The hardest part of all of this is IVTC'ing. Automated methods really don't cut it, especially if you plan to sync to GOUT. Darth Mallwalker is the IVTC master around here, so I think he'd be more qualified to help out with this particular aspect of your preservation(s). I don't know if he'd want to, so it'd be better to ask him.

Antcu, can I count on your help if (when) needed, for what concerns AviSynth and/or VirtualDub scripts and settings?

Yes :-) My real name is Anthony, BTW.

A picture is worth a thousand words. Post 102 is worth more.

I’m late to the party, but I think this is the best song. Enjoy!

—Teams Jetrell Fo 1, Jetrell Fo 2, and Jetrell Fo 3

Author
Time

althor1138 said:


If you have several captures you could try interpolating the frames with MflowFPS to remove dropouts etc.  Since each frame is identical except for the fact that it is from another capture it should only average out noise.


How would MflowFPS help?

DE

Author
Time

Anthony, thanks for all the info and tips.

But, my captures will never need IVTC as they are PAL, and if I'll resize to anamorphic, it will be PAL... or, if NTSC, I'll just change the framerate from 25 to 23,976 - thought also about this eventual change to use NTSC audio soundtracks and come back to the original film framerate - almost, as film is 24fps.

About DVD resolution, if I capture at 720 then crop at 704 (also if just 2 pixels are black on the sides it doesn't matter), this is a resolution accepted by the DVD standard, and I don't need to resize in the horizontal direction... at least, avoiding the horizontal resize will result in less interpolation, and this is good.

So, following my idea:

  1. Capture at 720x576
  2. Crop 8 pixels per side, obtaining 704x576
  3. Resize vertically from 576 to 768
  4. Crop 96 pixels from top and 96 pixels from bottom

 

At the end, I have a 704x576 25p 16/9, ready for DVD, with only vertical resize; if I want AVCHD, I could skip step 2 and obtain at the end 720x576 25p 16/9, with 8 pixels per side, without horizontal resize, OR add horizontal resize to setp 3. The original picture, without black borders, if 2.35:1 is 326 pixels high, then after the conversion to anamorphic will be 434 pixels - PAL, of course.

...but there is a problem: I tested some 1997 SE captures - the german edition - and the picture is 342 pixels high - so not 2.35:1... then, what it is best to do? Resize at the correct ratio of 2.35:1 and obtain a picture 434 pixels high, or retain the original AR and obtain a picture 454 pixels high?

About MflowFPS: don't know this script, I'll try it. I add that, as I remember, the luma of captures of the three THX trilogies (spanish, german, french) were identical, while the chroma were different... may I could only compare luma, but the dropouts are colored, so... how to approach this task?

OK, other questions later, maybe about Huffyhuv YVY2 and Lagarigh YV12...

Sadly my projects are lost due to an HDD crash… 😦 | [Fundamental Collection] thread | blog.spoRv.com | fan preservation forum: fanres.com

Author
Time

OK, so you have 704x342 (after cropping all 4 black borders).

Considering PAL PAR, that gives us 768x342 square pixels, so the DAR is ~2.24. We want to maintain this DAR so that the anamorphically-encoded DVD video, when stretched on your widescreen TV, will have the same proportions as the source video.

Now, notice that 854 width that I was throwing around before. That's for NTSC. You get that by doing the following calculation: (480 * 16)/9 = ?.

For PAL, we do a similar calculation: (576 * 16)/9 = 1024. So we now know that when the video is displayed on your widescreen TV, it will have an approximate resolution of 1024x576.

Using the same method as before, we need to adjust the height. The calculation is (342 / 768) * 1024 = 456. So, you need to make the size 720x456 (before adding black borders) in order for it to look correct on your widescreen TV after the anamorphically-encoded DVD video is stretched.

(Notice that 1024/456 = ~2.24! This is our DAR!)

How much do you need to crop from the top to get 342?

I can write the correct code for you if you get me that value.

A picture is worth a thousand words. Post 102 is worth more.

I’m late to the party, but I think this is the best song. Enjoy!

—Teams Jetrell Fo 1, Jetrell Fo 2, and Jetrell Fo 3

Author
Time

Anthony, you are right, the value was 456 but I wrote 454 - my mistake! And for my previous post, I didn't get that you was talking about theory... I was SURE you were asking me to REALLY upscale the picture to 768! Sorry! Still friends?

I appreciated also your explanation, but I love simple things...

So, if I need to resize any PAL picture (with an AR higher than 1:78) from 4:3 to 16:9, I just resize at 768 pixel height and crop 96 pixels from top and 96 pixels from bottom. Is that easy! ;-)

Maybe AviSynth takes some time more to resize my way, but I have no "difficult" math to remember, and find the exact crop needed... do you agree?

Still not sure to do the anamorphic conversion... "the less resize, the better", my grampa used to say :) but maybe he was wrong, he said that long time ago...

 

At the end, someone else has noted that the 1997 SE aspect ratio was NOT 2.35:1? Or it is, and only the german version is 2.24:1?

Sadly my projects are lost due to an HDD crash… 😦 | [Fundamental Collection] thread | blog.spoRv.com | fan preservation forum: fanres.com