pixels size and questions

[gary]

so on this picture how can i have tiny lines on the right ? They are half a pixel high in 48 ( so 1 pixel in 96 )

Well, as you can see, 31 out of 32 are fine so it is NOT a resolution problem, now this *could* be a genuine bug, but without someway of reproducing it (i.e. the original video input, version of Video2NBTV, format used) I can't be sure of the cause, nor can I fix it if it is a Video2NBTV bug. So if you can send me that information I can have a look at it. However, it is probably a boundary condition associated with the resizing algorithm, something I don't actually have control over, but it should be corrected by applying the resizing rules indicated above. None-the-less I would like to investigate it so if you can please send me the original video or let me know the procedure used to create it.

[Lowtone]

Hello Gary, thanks for your long reply. I'm not sure to understant all the technical aspects because of my poor knowledge of those things.

Well I don't understand the width distortion because because 32x48 and 64x96 have the same aspect ratio, but yes, if you want to exceed the NBTVA bandwidth specification (9.6kHz) then the optimal video size would be 80x120 (for 48kHz).

Why is it 80 ? I took 64 because is it twice 32 and it was easy for me to make tescards and things.

Note that if you use the Brown/Robers colour format (i.e. the current RELEASE version of video2NBTV) then the overall resolution WILL be constrained by the 10kHz smoothing filter (effectively 32x48) because it produces an NBTVA standard format.

I don't know this color format
is 10㎑ enough for color ?

Not also though, that ideal size of the video would be 294x441 as the image will be resized to that anyway during conversion.

The picture is getting bigger, and closer to CRT television. It would make big files. I think i need to try all those different sizes an see what makes the best picture.

Well, that will only work (for B/R colour format) IF the resizing algorithm used by Premiere is BETTER than that used by Video2NBTV (possible but unlikely) AND, and this is a big AND, the size of the image you produce is 294x441 because, whatever you put into it WILL be converted to that size prior to conversion. If it isn't then Video2NBTV's resizing algorithm will be used anyway, therefore removing any benefit gained by resizing with Premiere.

The pictures are quite good, for movies, and everything natural. But if make testcards or abstract drawings with straght lines and squares, it gets a little blurry on the edges.

If you are using one of the Alpha release versions then the ideal size will be 80x120 because in that case no resizing will need to be done at all in the vertical direction (the image will converted to 32 x 120) and so the pixels in the vertical direction will be passed one for one to the NBTV video.

I'm not sure to understand. 32×120 makes a very long picture ( and distorted i guess ) So that means there are more pixels than 48, so the pixels along the lines don't have the same size as the lines ( = non square pixels )
That is the "defect" that i was trying to use in order to increase the quality ( according or not to the NBTVA format, if the thing is working fine, well it is good for me )
But non-square pixels can make some odd effect.
In a movie, if there is a zoom, or travelling on rails, as the pixels aren't square, it seems that it zooms faster in one direction than the other one.
Quite amusing though.

It is *probably* a good idea to do that. Video2NBTV asks DirectShow to do the conversion from the videos native frame rate to 12.5 - I suspect that DirectShow doesn't do a particularly good job of that. If the frame rate presented to the NBTV conversion algorithm is NOT 12.5 then it will use a simple drop frame or duplicate frame algorithm to compensate.

I don't know, i've never see any problem with the frame rate.
The only problem i could see is if i want to take the AVI and convert it to another standard that use more fps. It could work, but won't be the best thing ( like super 8 to VHS from 18 to 25 fps, it makes weird things while looking frame by frame )

Well it *should* be given the constraints given above.

mmm ok i will experiment further

Well, as you can see, 31 out of 32 are fine so it is NOT a resolution problem,

If you wan i can make ALL the lines like this one !
I even done new testcards to see if i could go further !
And i open this thread because i couldn't !
I thought the problem was the 48㎑. So i wanted to increase everything.

now this *could* be a genuine bug,

Ah yes
And i wonder how many videos were made like this before we noticed

but without someway of reproducing it (i.e. the original video input, version of Video2NBTV, format used) I can't be sure of the cause, nor can I fix it if it is a Video2NBTV bug.

V2 alpha, NBTVA format.
If you fix it i would keep the version with bug because i found it good. But i understand that you must be as close as possible to the club standard.

So if you can send me that information I can have a look at it. However, it is probably a boundary condition associated with the resizing algorithm, something I don't actually have control over, but it should be corrected by applying the resizing rules indicated above. None-the-less I would like to investigate it so if you can please send me the original video or let me know the procedure used to create it.

File was an AVI in 64×96 12.5p square pixels, color.
It was a big size file but i could post pictures. or maybe make a short file.
I just put the fil in the software and click. It makes a WAV file. Everything went fine.

Here are some pictures at the size 64×96

[gary]

Hello Gary, thanks for your long reply.

Oh no problem, I am always very happy to receive feedback about my software (as, I am sure, does Dominic). If there is a bug I WANT to hear about it and will always attempt to fix it if it is possible, there should be no need for workarounds if it is working properly. NBTV is always about sharing whether it be ideas, comments, bug reports etc)

Why is it 80 ? I took 64 because is it twice 32 and it was easy for me to make tescards and things.

This is simply because for a 48kHz signal we always have 120 samples to represent a line of NBTV video (for full bandwidth) - to maintain an aspect ratio of 3:2 that implies that the horizontal resolution be 32 * 120 / 48 = 80. This image (80 x 120) is then resampled (by Video2NBTV) to 32 x 120 to give, at 12.5 frames per second a sample rate of 48kHz - so note, in principle, we have a resolution of 120 pixels in the vertical direction for this format (i.e. 24kHz bandwidth).

I don't know this color format
is 10㎑ enough for color ?

Well this is simply the format used by the RELEASE version of Video2NBTV and is a format devised by Vic Brown and Klaas Robers as a potential backwards compatible NBTVA standard for colour NBTV video.

Yes and no, 10kHz is enough for the luminance (Y) value of this format (hence the backwards compatibility to NBTVA B&W format) - the colour information is actually stored digitally, and needs to be decoded digitally, which, in essence, means the overall signal is using a bandwidth of 44.1 x 16 kHz if it were to be transmitted serially. Thus the *colour* format is actually using the full bandwidth of the 44.1kHz signal.

The picture is getting bigger, and closer to CRT television. It would make big files. I think i need to try all those different sizes an see what makes the best picture.

Yes, the original video should be as high as possible, but preferably the resolution given above - anything else could result in a lower resolution NBTV video. Even going from a HIGHER resoultion to a LOWER resolution can result in quality loss due to the necessity for imperfect low pass filtering.

The pictures are quite good, for movies, and everything natural. But if make testcards or abstract drawings with straght lines and squares, it gets a little blurry on the edges.

Yes, this is a common problem with the algorithms used for image resizing and derives from the fact that there may not be an integral number of pixels surrounding the target pixel. With higher resolution images this is barely noticeable, but in low resolution images it has an effect - BUT THIS SHOULD NOT BE PERCEPTIBLE ON A MECHANICAL MONITOR.

I'm not sure to understand. 32×120 makes a very long picture ( and distorted i guess ) So that means there are more pixels than 48, so the pixels along the lines don't have the same size as the lines ( = non square pixels )

Yes it would if that is the way the picture is finally reproduced. In practice, the digital signal is converted into an analogue signal and the 120 pixels are interpolated into 48 pixels (this is a DIGITAL analogy, in practice a mechanical monitor makes much better use of this additional resolution than any other form of monitor ever produced including CRTs LCDs, plasma etc) because the resolution it can display is only limited by aperture distortion which in turn is inversely proportional (in terms of picture brightness) to the square of the modulated light intensity - debate expected).

In other words, the 32x120 image is a distorted long picture (for NBTVA format) but is corrected geometrically by the monitor display device).

That is the "defect" that i was trying to use in order to increase the quality ( according or not to the NBTVA format, if the thing is working fine, well it is good for me )
But non-square pixels can make some odd effect.
In a movie, if there is a zoom, or travelling on rails, as the pixels aren't square, it seems that it zooms faster in one direction than the other one.
Quite amusing though.

Well if the picture is being interpolated correctly then there should be no artifacts. Each frame of video is processed individually so movement or zooming should have no effect at all on the converted picture.

I don't know, i've never see any problem with the frame rate.
The only problem i could see is if i want to take the AVI and convert it to another standard that use more fps. It could work, but won't be the best thing ( like super 8 to VHS from 18 to 25 fps, it makes weird things while looking frame by frame )

Well that's good - all I am saying is that is the one area where I can see possible improvement over what Video2NBTV is doing.

If you wan i can make ALL the lines like this one !
I even done new testcards to see if i could go further !
And i open this thread because i couldn't !
I thought the problem was the 48㎑. So i wanted to increase everything.

Fine, but I need an input file so I can reproduce it. There should NOT be any perceptible problem for a 32 x 48 image.

If you fix it i would keep the version with bug because i found it good. But i understand that you must be as close as possible to the club standard.

Don't worry, I would accommodate your requirements in the "fix"

[gary]

If you fix it i would keep the version with bug because i found it good. But i understand that you must be as close as possible to the club standard.

Actually with the alpha version I don't - it will include the club standard but also many other "variants" useful for experimentation.

File was an AVI in 64×96 12.5p square pixels, color.
It was a big size file but i could post pictures. or maybe make a short file.
I just put the fil in the software and click. It makes a WAV file. Everything went fine.

Here are some pictures at the size 64×96

Yes, but can you post the one that made the error in your previous post?