What's the difference from resizing to 640x480 before cropping and after?

[CZbwoi]

So the original captured source comes out of Vdub as 740x480, but it should be 640x480 as the tape is a Hi8 tape.




If I insert this code into AvsPmod (avisynth) I get this, and from thereon when I open up the script in MeGUI it says it's 4:3 (1.333333):

crop(6,0,-10,-6)
Spline36Resize(640,480)

It looks pretty right to me.


But if I insert the resize first and press enter, the video slightly shrinks in the preview..yet still looks pretty much the same from what I see (unless they're not?) :

Spline36Resize(640,480)
crop(6,0,-10,-6)

Opening this up in MeGUI it tells me that its 104:79 (1.316). When I change the DAR to 4:3 (1.333333) it shrinks the video vertically slightly, changing it from what I ended up with after resizing, cropping and getting the exact size (width and height) it should be at. It's not how it should be anymore, even if slightly vertically, I don't want it to be altered. But from what I was told, I know that I have to pick a DAR like one of those after cropping to weird or uneven sides due to uneven black bars and whatnot. But it's resizing the video slightly when I do so.

Which one is the right method to use?


Also, in MeGUI I see 2 options for 4:3. It's defaulted at 4:3 (1.333333) when I insert the first method's code but there's another option called ITU 4:3 NTSC (1.367377). When I select the latter it slightly shrinks the videos vertically even more; is this the one that's more true to the source, should I use it or no? What's the difference here?

[Bernix]

If you crop after resize, you dont get the right size. So picture is smaller than 640x480 so less detailed.
And I think, that better downscaled filter is Lancsoz or bicubic.
I am noob, but when nobody answered you

[KarMa]

You cropped off the tape noise at the bottom and then stretched the video back to 480 pixel, which causes unneeded quality loss and aspect ratio errors. So it would be better to crop off the tape noise and then replace it with black bars, or just crop it off and leave it at 474 pixels.


The script I used (minus the YV12 conversion I had to use with this PNG picture)

crop(6, 0, -12, -6) #Sum of 18 horizontal crop
Spline64Resize(640,474) # 
AddBorders(0,0,0,6) #this adds 6 pixels to the bottom to return the picture to 640x480

or you can

Spline64Resize(656,480) # See here I am resizing to ITU 4:3 NTSC (1.367) which is the proper aspect ratio with analog sources and most DVDs.     
crop(6, 0, -10, -6) #Sum of 16 instead of 18 horizontal crop, will give you 640x474
AddBorders(0,0,0,6) #Returns the picture to 640x480

Both scripts should give you very nearly the same results, at least in my own comparisons on my own videos.

You can take out the addboarders, but will be left with 640x474. Which is fine for personal use, but DVDAuthoring tools and Youtube might stretch it to 640x480 or just down convert to something like 360p. So that's its purpose, useful filler.

I don't know if Spline64Resize is the best resizer for this job so that certainly can be changed if needed.

But if I insert the resize first and press enter, the video slightly shrinks in the preview..yet still looks pretty much the same from what I see (unless they're not?) :

Spline36Resize(640,480)
crop(6,0,-10,-6)

You can resize first, but you have do 656x480 (480 X 1.367 = 656.16) The ITU 4:3 NTSC preset is in MeGUI.

[CZbwoi]

Spline64Resize(656,480) # See here I am resizing to ITU 4:3 NTSC (1.367)

Wait a minute, you're telling me that all of my tapes should be in 656x480 instead? Is that (656x480) the right size for just the actual footage or the entire capture from the device including the bars and tape noise?

All of my American Hi8/8mm/VHS/VHS-C tapes? Everybody's been telling me that they should be in 640x480 til now...that's been wrong all along?

You can take out the addboarders, but will be left with 640x474. Which is fine for personal use, but DVDAuthoring tools and Youtube might stretch it to 640x480 or just down convert to something like 360p. So that's its purpose, useful filler.

If these were just for media player playback, Google Drive playback (which uses the YT player), phone playback, TV/Blu-ray player USB playback, would you say it'd be best to add those black borders and then make it 640x480 or no? Do you have any personal experience with home tapes?

I don't know if Spline64Resize is the best resizer for this job so that certainly can be changed if needed.

The only other options would be 12resize and 36, right?

You can resize first, but you have do 656x480 (480 X 1.367 = 656.16) The ITU 4:3 NTSC preset is in MeGUI.

So resize to that first, crop, insert the ITU 4:3 NTSC preset. What if I crop weirdly and uneven? I see you mention Sum of 18 horizontal crop and whatnot, what if my crops aren't even because one side has slightly more black or something, does that alter anything or resize anything when selecting the ITU preset?

[manono]

Wait a minute, you're telling me that all of my tapes should be in 656x480 instead?

If you resize directly from the 720x480 source, yes.

Everybody's been telling me that they should be in 640x480 til now...that's been wrong all along?

Typically you might crop a total of 16 columns of pixels (8 from both left and right, or 6 and 10 in one of KarMa's scripts) to leave 704x480. And that's what gets resized to 640x480. With other crops, and if you're cropping away the noise at the bottom, then you're left with different dimensions to resize and you might wind up with a different resolution if trying to keep the aspect ratio. This is one reason why KarMa suggested replacing the junk at the bottom with black - to lessen the amount of the resize, or to degrade the video a bit less from a resize, or to make it possible to use the common 640x480 resize. Because of overscan, there's a good chance you won't see that black when viewing on a television screen.

[jagabo]

Spline64Resize(656,480) # See here I am resizing to ITU 4:3 NTSC (1.367)

Wait a minute, you're telling me that all of my tapes should be in 656x480 instead?

That was explained to you before:
https://forum.videohelp.com/threads/378101-My-Premiere-Pro-exports-give-me-little-black...=1#post2441702
And I'm pretty sure in a few other posts.

[CZbwoi]

Spline64Resize(656,480) # See here I am resizing to ITU 4:3 NTSC (1.367)

Wait a minute, you're telling me that all of my tapes should be in 656x480 instead?

That was explained to you before:
https://forum.videohelp.com/threads/378101-My-Premiere-Pro-exports-give-me-little-black...=1#post2441702
And I'm pretty sure in a few other posts.

Nowhere in that post or page does it say 656x480 is for analog tapes, and that whole thread I made was about a widescreen camcorder's (DV?) footage, not analog tape sources like this one. How am I supposed to take information about this from post #50 in a thread where I was only thinking about how to get rid of borders from an encoded DV source, when even now I still barely see a vague correlation with the linked post?

I just wanted to know what analog tapes should've been resized to all this time directly after capture or for the final resolution, everyone's been dropping 640x480 on me before KarMa just told me it's 656x480, unless I forgot someone specifically saying that.

Typically you might crop a total of 16 columns of pixels (8 from both left and right, or 6 and 10 in one of KarMa's scripts) to leave 704x480. And that's what gets resized to 640x480.

Why wouldn't it be a good idea to resize to 656x480 over here as the final size if that's the proper aspect ratio for analog sources? Do the extra 16 pixels in 656 always only account for the black added areas?

With other crops, and if you're cropping away the noise at the bottom, then you're left with different dimensions to resize and you might wind up with a different resolution if trying to keep the aspect ratio.

Doesn't everyone capturing analog tapes have this problem of cropping noise at the bottom and are left with unnatural dimensions? Do all of you resort to adding black borders at the bottom to compensate? I've just never heard of adding borders to fix something when I'm taught to crop them out. But hey, if that's normal for restoring videos...then what do I know. Is this what all the pros do?

[_Al_]

I think you overdo stuff, just do this what was recomended and do not think what is better or worse,
take away those 6 and 10 from sides, remove bottom BUT just pad it right back, black color to not mess up vertical resolution. I recomended that padding in one post somewhere before as well

input is 720x480 so:

crop(6, 0, -10, -6)
AddBorders(0,0,0,6)
Spline36Resize(640,480)


most of us resizes from 720x480 directly to 640x480 and everything is fine and dandy, not thinking about that 720 vs. 704 hell. But mostly it is discussed with DVavi in mind, camcorder capture, not VHS. Someone said that it depends on particular camcorder if correct aspect ratio is in 704 or in whole 720, not sure what's correct. I shot video with VX2000 DV camcorder to 720x480, lot's of weddings and was getting square pixel right from those 720, not 704 (after cropping). Nobody noticed anything, complained or something. But working with VHS I'd do what they say, aspect ratio is correct for 704, so you resize that to 640, leaving vertical resolution untouched for quality reasons.

[manono]

Why wouldn't it be a good idea to resize to 656x480 over here as the final size if that's the proper aspect ratio for analog sources?

If you want to keep the full original 720x480, then you might do it that way.

Do the extra 16 pixels in 656 always only account for the black added areas?

It's not always 16 columns of black pixels within that 720 width. I've seen more and I've seen less. I've seen DVDs with no black on the sides at all. As you may or may not know, old films used Academy Ratio and should be resized to a roughly 1.37:1 ratio such as 656x480. DVDs of films from that era often reflect the intended aspect ratio by having no black on the sides.

Doesn't everyone capturing analog tapes have this problem of cropping noise at the bottom and are left with unnatural dimensions?

Often, yes.

Do all of you resort to adding black borders at the bottom to compensate?

Some do, some don't. I usually crop the head-switching noise away. I have added black bars above and below, though. Once done cropping away the bottom noise, I'll try and split the added black borders to both above and below rather than have it all below. It depends on how much is cropped and your own feelings about it. Since we're only talking about a very few percent of possible aspect error, it's not as if you'll ruin your viewing experience by doing a resize that's slightly off. You could even crop into the active video to minimize the aspect error when resizing, although one doesn't really like to have to do that.

[jagabo]

Spline64Resize(656,480) # See here I am resizing to ITU 4:3 NTSC (1.367)

Wait a minute, you're telling me that all of my tapes should be in 656x480 instead?

That was explained to you before:
https://forum.videohelp.com/threads/378101-My-Premiere-Pro-exports-give-me-little-black...=1#post2441702
And I'm pretty sure in a few other posts.

Nowhere in that post or page does it say 656x480 is for analog tapes, and that whole thread I made was about a widescreen camcorder's (DV?) footage, not analog tape sources like this one. How am I supposed to take information about this from post #50 in a thread where I was only thinking about how to get rid of borders from an encoded DV source, when even now I still barely see a vague correlation with the linked post?

That was just the first one I found. Here is another:

https://forum.videohelp.com/threads/378055-I-need-some-help-converting-my-old-VHS-VHS-C...=1#post2441938

And several posts before that discuss the same issue.

In an ITU cap (virtually every modern capture device follows the ITU spec) the 4:3 picture is in a 704x480 frame. If the frame is 720 pixels wide those extra 16 pixels are just padding (so you don't lose any of the actual picture when the cap is slightly off center -- as was the case in your elgato vs. happuage thread). So the 720x480 frame represents something slightly wider than 4:3. If you crop to 704x480 then you want to resize to 640x480. If you keep the full 720 pixels of the source correct frame width is about 656 pixels.

[CZbwoi]

Thanks for the input _AL_

Nobody noticed anything, complained or something. But working with VHS I'd do what they say, aspect ratio is correct for 704, so you resize that to 640, leaving vertical resolution untouched for quality reasons.

Well, not really untouched as I have to get rid of the tape noise at the bottom, so I'd have to pad it at the top/bottom which doesn't look pleasant or resize up which will stretch it out...

Some do, some don't. I usually crop the head-switching noise away. I have added black bars above and below, though. Once done cropping away the bottom noise, I'll try and split the added black borders to both above and below rather than have it all below. It depends on how much is cropped and your own feelings about it. Since we're only talking about a very few percent of possible aspect error, it's not as if you'll ruin your viewing experience by doing a resize that's slightly off. You could even crop into the active video to minimize the aspect error when resizing, although one doesn't really like to have to do that.

Thanks for telling me your preferences. The thing is I don't want the aspect to be even *slightly* off, I want it to be the natural and pure size of how the people were to the centimeter, surely I'm not the only one that feels that way especially about family footage lol.

That was just the first one I found. Here is another:

https://forum.videohelp.com/threads/378055-I-need-some-help-converting-my-old-VHS-VHS-C...=1#post2441938

And several posts before that discuss the same issue.

In an ITU cap (virtually every modern capture device follows the ITU spec) the 4:3 picture is in a 704x480 frame. If the frame is 720 pixels wide those extra 16 pixels are just padding (so you don't lose any of the actual picture when the cap is slightly off center -- as was the case in your elgato vs. happuage thread).

Yes, I know we've discussed this topic and extensive terminology before a few times but nobody mentioned 656x480 being the right actual source aspect for a pure resize, that's what I'm saying. I needed it laid out to me flat today with those actual 2 numbers, if you mentioned it by pointing out how everything works in similar frames and ratios without ever actually saying "656x480" is the pure capture source resize then it flew over my little head trying to grasp all this. That's what I'm referring to.

If you crop to 704x480 then you want to resize to 640x480.

But with analog tape capture sources you'll always have tape noise at the bottom so it'd make no sense to ever keep it at 480 by cropping to 704x480, right? Unless you want tape noise fluttering at the bottom for some vintage reason.

May I ask what you yourself do to get rid of the tape noise and your steps after that? Do you crop it out and then pad with black at the bottom back up to 480? Do you split padding at both the top and bottom? Do you resize it up to 480 and alter/stretch out the image slightly? Something else to get rid of the tape noise like blurring it with a filter?

[ackboo]

This doc might help, analog video aspect ratio is defined by the pixel aspect ratio. For NTSC it is 10:11, this will give the proper appearance to the content of your video.

How do I Convert Between Square and Non-Square?
http://www.lurkertech.com/lg/pixelaspect/

In any case you should not resize first to 640x480, that's wrong and adding complications for nothing. Crop first to preserve your source and then resize or keep anamorphic.

In your case I would deinterlace, crop to 704x472, addborder to restore to 704x480 (4 up 4 down) and resize to 640x480, because 704*10/11=640
Or to preserve quality, keep the video anamorphic and deinterlace, crop to 704x472, addborder to restore to 704x480, add --sar 10:11 to the x264 parameters to make the video display as 640x480

Another option to avoid borders and get rid of the head switching noise is to deinterlace, crop to 692x472, add --sar 10:11 to the x264 parameters, this should display as 629x472 AR=1,332

You can also crop and upsize to 640x480 at the cost of some quality

But then I'm only a noob here. Still trying to wrap my head around all this.

[lordsmurf]

This is why to need to mask, never crop.
It's too confusing, and usually unnecessary since everything but computers crop.
Even for Youtube, I often mask, because I don't want to stretch pixels.

You're doing too much to your videos, and that often leads to problems.
With VHS source, you have enough as is!

[jagabo]

Yes, I know we've discussed this topic and extensive terminology before a few times but nobody mentioned 656x480 being the right actual source aspect for a pure resize, that's what I'm saying.

You should have been able to figure that out for yourself from the information given. It's trivially obvious.

If the 4:3 image is in a 704x480 sub-frame then the full 720x480 frame must be wider than 4:3. How much? A factor of 720/704 so instead of 640x480 you need a width of 640 * 720 / 704 = 654.5454... But you can't have partial pixels so you have to round up or down. 655 is out because you can't encode YV12 video with odd frame sizes. 654 is slightly more accurate (than 656) but its mod2 and should therefore be avoided. 656 is mod16, a good width for all modern codecs.

Or you could have done the math:

DAR = FAR * SAR
DAR = 720:480 * 10:11
DAR = 720 / 480 * 10 / 11
DAR = 1.363636...

480 * 1.363636... = 654.5454...

And finally, all this is wrong because the commonly used SAR values are all approximations and are off by about 0.1 percent.

http://xpt.sourceforge.net/techdocs/media/video/dvd/dvd04-DVDAuthoringSpecwise/ar01s02.html

And there's the problem of aspect ratios defined for DVD (really the MPEG 2 spec) where the 4:3 image is contained in the full 720x480 frame, and the ITU spec where the 4:3 image is in a 704x480 sub-frame, usually ignored by the industry.

May I ask what you yourself do to get rid of the tape noise and your steps after that?

What I do depends on the final destination. I might Crop() then AddBorders() to restore the frame size when I need a particular frame size -- like when making DVDs. For my own use I usually crop away black borders and otherwise leave the frame alone, encoding to h.264 with SAR flags. All my players respond to SAR flags. If I needed a video for players that don't respond to SAR flags I'll crop and resize to the nearest mod4 square pixel resolution.

[johnmeyer]

I never re-size because re-sizing is always going to degrade the video. I just crop the head noise with a little small black strip. If there is a little black around the image, no one is going to care because so much video you watch on TV has black bars of some sort, from 2:35 films shown at 16:9 to 4:3 video shown on widescreen TVs, etc.

Cropping also has the advantage of being faster, and it can usually be done in one command, whereas re-sizing is often a two-step operation.

[jagabo]

All of this has been discussed before in CZbwoi's threads.

[Budman1]

Wouldn't just setting the Display aspect work as well?

crop(6,0,-10,-6)
Spline36Resize(640,480)
dar=1.485
return last

[CZbwoi]

Thanks, ackboo.

This is why to need to mask, never crop.
It's too confusing, and usually unnecessary since everything but computers crop.
Even for Youtube, I often mask, because I don't want to stretch pixels.

You're doing too much to your videos, and that often leads to problems.
With VHS source, you have enough as is!

What do you mean by mask and how do you do that? This isn't for VHS sources as of now btw, just for my Hi8 tapes since...I'm still waiting on you and your team to give me an answer about capturing my VHS tapes, you said it'd be a day or two for a response back and it's been over a week...did you ever get anything? :/ Do you still actually do it? I'm assuming you're the same lordsmurf I talked to over there lol.

What I do depends on the final destination. I might Crop() then AddBorders() to restore the frame size when I need a particular frame size -- like when making DVDs. For my own use I usually crop away black borders and otherwise leave the frame alone, encoding to h.264 with SAR flags. All my players respond to SAR flags. If I needed a video for players that don't respond to SAR flags I'll crop and resize to the nearest mod4 square pixel resolution.

Thanks. I have to figure out if the YouTube player in Google Drive responds to SAR flags.

I never re-size because re-sizing is always going to degrade the video. I just crop the head noise with a little small black strip. If there is a little black around the image, no one is going to care because so much video you watch on TV has black bars of some sort, from 2:35 films shown at 16:9 to 4:3 video shown on widescreen TVs, etc.

Cropping also has the advantage of being faster, and it can usually be done in one command, whereas re-sizing is often a two-step operation.

Thanks, John. The reason why I care about cropping the black around the image is because they aren't even as you can see, hence the 6 and -10. if I leave it as is then the video would be slightly off center on the screen since there's more black on the right side. There's also a strip of green tape noise on the right and I think it'd be best to crop that out too, and to get to that you have to get rid of the black first, unless there's a method to get rid of that green strip I don't know about?

And your final sentence, isn't cropping the head noise with a little small black strip also a two-step operation? Crop() and AddBorders() is two steps, no?


Budman1: Regarding your code, since we're cropping from the bottom to get rid of the noise and cutting the natural capture length size of 480, resizing to 640x480 would stretch out the length and lose quality.

[manono]

And your final sentence, isn't cropping the head noise with a little small black strip also a two-step operation? Crop() and AddBorders() is two steps, no?

But Crop alone isn't, which is what he said. He was saying he doesn't bother to cover it up or resize - just Crop it away. Yes, Crop/AddBorders is, but so what? If you prefer, you can do it in a single step with the Letterbox command. That is:

Crop(0,0,0,-10).AddBorders(0,0,0,10)

is the same as:

Letterbox(0,10)

What do you mean by mask and how do you do that?

He means 'cover it up' and you do it with Crop/AddBorders or with Letterbox.

[CZbwoi]

And your final sentence, isn't cropping the head noise with a little small black strip also a two-step operation? Crop() and AddBorders() is two steps, no?

It is, but so what? If you prefer, you can do it in a single step with the Letterbox command. That is:

Crop(0,0,0,-10).AddBorders(0,0,0,10)

is the same as:

Letterbox(0,10)

What do you mean by mask and how do you do that?

By Crop/AddBorders or by Letterbox.

Aha, I didn't know of that command, thanks for letting me know.

[lordsmurf]

I just crop the head noise with a little small black strip.

You're not cropping, you're masking.
Using the wrong term is why we always run into issues with video.

You know what you mean, and I know what you mean -- but newbies do not.

The problem is compounded because a lot of software uses the wrong terms, too. Premiere uses "clipping", while Avisynth uses the "crop" tool to mask, and VirtualDub uses both "crop" and/or "resize" tools to mask.

[jagabo]

VirtualDub has a rectangular Fill filter that can be used for masking. You could always do something similar in AviSynth by making a black video and using Overlay() (but that would be a waste of time and effort).

Worrying about whether it takes one step or two is silly. Simple filters like Crop(), AddBorders(), and Letterbox() are very fast compared to h.264 encoding. Performing the operation in one step rather than two might take a few seconds off an hour long encode.

[jagabo]

Wouldn't just setting the Display aspect work as well?

You can use DAR flags with codecs or containers that support it. But then you need to calculate what the DAR should be. When you use SAR flags (with no resizing) all you need to know is the SAR of the source. Cropping and padding don't effect the SAR.

[manono]

I just crop the head noise with a little small black strip.

You're not cropping, you're masking.
Using the wrong term is why we always run into issues with video.

Funny, I read the same thing you did and it didn't register. You're right, of course, and thanks for the gentle correction. Even though that sentence was at the beginning of the post, the one I was commenting on was later:

Cropping also has the advantage of being faster, and it can usually be done in one command, whereas re-sizing is often a two-step operation.

Not that it makes much difference whether covering with black is done in one step or two (or even not at all), which is what CZbwoi was asking. And I usually both crop and resize and don't bother masking. I work with VHS tapes and the amount of black on left and right sometimes totals more than 16 columns of pixels so it's possible to crop and resize at the same time keeping the AR for eventual reencode for DVD. In addition, any quality loss is minimal compared to the crap quality of the sources to begin with. Or, even with a normal amount of black one can crop the head-switching noise and resize to 720x480 rather than 704x480, and keep the AR.