Soccer matches - CBR vs VBR decision

[Santuzzu]

I have a dilemma about a video quality for soccer matches I recently downloaded. I prefer to download videos with constant 60fps, but bitrate is problematic because I'm not sure which mode brings a better quality. Here is the screenshot of the bitrate viewer output for two videos of the first half of the same soccer match.



On the left side is a CBR video and VBR is on the right side. Which one do you think is of a better quality? I would also like to read a more detailed analysis to be able to make decisions for other videos.

[davexnet]

Nobody can tell anything from those graphs that's particularly useful. Is this H.264/Hevc encoding? CRF encoding is usually better.
Why are you re-encoding your download?

[Santuzzu]

This is H.264 encoding. I'm not re-encoding anything, I downloaded these two video files and want to keep one but I'm not sure which one is of a better quality.

[davexnet]

Can you post a 30 second clip from each one so we can take a look at it ?

[Santuzzu]

Here:

https://www.mediafire.com/file/2k8ij81un5casm3/CBR.mkv/file
https://www.mediafire.com/file/eo0myduprj789iy/VBR.mkv/file

[_Al_]

you try to compare two different TV footages basically
,
it is presented with offset, CBR is 69.5 frames behind , and it is not accurate because of deinterlacing or diffeent broadcast, there is 2pixel offset vertically


anyway both have pretty much enough bitrate for what you do,
do whatever works for you, CBR either wastes or has not enough bitrate, image could have variable quality, VBR tries to be balanced with whatever it has

EDIT: for 2pass VBR, you might use just one pass VBR, which could be off as well

[jagabo]

The differences aren't because one is CBR and the other is VBR. VBR has less small low contrast detail (in the grass for example) probably from a noise reducer or multiple re-encodings. It also has poorer deinterlacing (the original sources was 25i, 50 fields per second, and both were converted to 60p with field blending and deinterlacing). Overall CBR is probably better but I would guess that the VBR video looks better at those big bitrate peaks.

[Santuzzu]

The differences aren't because one is CBR and the other is VBR. VBR has less small low contrast detail (in the grass for example) probably from a noise reducer or multiple re-encodings. It also has poorer deinterlacing (the original sources was 25i, 50 fields per second, and both were converted to 60p with field blending and deinterlacing). Overall CBR is probably better but I would guess that the VBR video looks better at those big bitrate peaks.

How can I check properties of the original source for any video? Is there any way to do that or you say it based on experience?

[badyu17]

CBR (TNT) easily looks better on this sample, VBR has more blocking, smearing and is less stable.

[davexnet]

The differences aren't because one is CBR and the other is VBR. VBR has less small low contrast detail (in the grass for example) probably from a noise reducer or multiple re-encodings. It also has poorer deinterlacing (the original sources was 25i, 50 fields per second, and both were converted to 60p with field blending and deinterlacing). Overall CBR is probably better but I would guess that the VBR video looks better at those big bitrate peaks.

How can I check properties of the original source for any video? Is there any way to do that or you say it based on experience?

I guess the "original source" in this case is what was captured off the air. Do you have this? Or do you mean your downloaded source?
You can open the files in mediainfo in text view to get a bunch of details, however in the case of these samples, not too much was revealed

[Santuzzu]

The differences aren't because one is CBR and the other is VBR. VBR has less small low contrast detail (in the grass for example) probably from a noise reducer or multiple re-encodings. It also has poorer deinterlacing (the original sources was 25i, 50 fields per second, and both were converted to 60p with field blending and deinterlacing). Overall CBR is probably better but I would guess that the VBR video looks better at those big bitrate peaks.

How can I check properties of the original source for any video? Is there any way to do that or you say it based on experience?

I guess the "original source" in this case is what was captured off the air. Do you have this? Or do you mean your downloaded source?
You can open the files in mediainfo in text view to get a bunch of details, however in the case of these samples, not too much was revealed

Yes, I assumed that "original source" is what was captured off the air, that's why I wanted to know if there is any way to find out what is the original source of any video. I guess it's better to download "raw" videos and process them later on your own, than to download already processed videos with visible artifacts.

[badyu17]

No the original source was 1080i25 because that's the standard in Europe. Since these are TV stations from US, the picture was already converted from 25i to 29.97i or 59.94p. Honestly I don't think these videos were converted with simple blending because TV stations today use much better converters with motion interpolation. The motion in these videos is good, not many artifacts from conversion.

[davexnet]

The differences aren't because one is CBR and the other is VBR. VBR has less small low contrast detail (in the grass for example) probably from a noise reducer or multiple re-encodings. It also has poorer deinterlacing (the original sources was 25i, 50 fields per second, and both were converted to 60p with field blending and deinterlacing). Overall CBR is probably better but I would guess that the VBR video looks better at those big bitrate peaks.

How can I check properties of the original source for any video? Is there any way to do that or you say it based on experience?

I guess the "original source" in this case is what was captured off the air. Do you have this? Or do you mean your downloaded source?
You can open the files in mediainfo in text view to get a bunch of details, however in the case of these samples, not too much was revealed

Yes, I assumed that "original source" is what was captured off the air, that's why I wanted to know if there is any way to find out what is the original source of any video. I guess it's better to download "raw" videos and process them later on your own, than to download already processed videos with visible artifacts.

With a little experience it is usually possible to find out things like original framerate, as jagabo pointed out.
I never noticed it my self, when I looked at it in Virtualdub2 frame by frame it looked like 60 unique field,
I never noticed any bending. I was looking at the players and the ball, they're so small - perhaps it's my eyes!

Unless you're lucky enough to be able to capture the off air stream without loss yourself, I don't know how you would get such a thing

[Santuzzu]

The differences aren't because one is CBR and the other is VBR. VBR has less small low contrast detail (in the grass for example) probably from a noise reducer or multiple re-encodings. It also has poorer deinterlacing (the original sources was 25i, 50 fields per second, and both were converted to 60p with field blending and deinterlacing). Overall CBR is probably better but I would guess that the VBR video looks better at those big bitrate peaks.

How can I check properties of the original source for any video? Is there any way to do that or you say it based on experience?

I guess the "original source" in this case is what was captured off the air. Do you have this? Or do you mean your downloaded source?
You can open the files in mediainfo in text view to get a bunch of details, however in the case of these samples, not too much was revealed

Yes, I assumed that "original source" is what was captured off the air, that's why I wanted to know if there is any way to find out what is the original source of any video. I guess it's better to download "raw" videos and process them later on your own, than to download already processed videos with visible artifacts.

With a little experience it is usually possible to find out things like original framerate, as jagabo pointed out.
I never noticed it my self, when I looked at it in Virtualdub2 frame by frame it looked like 60 unique field,
I never noticed any bending. I was looking at the players and the ball, they're so small - perhaps it's my eyes!

Unless you're lucky enough to be able to capture the off air stream without loss yourself, I don't know how you would get such a thing

BTW, can someone explain the whole process of TV broadcasting? I mean, the whole process starting from acquiring video using cameras to recording it on PC.

[jagabo]

The differences aren't because one is CBR and the other is VBR. VBR has less small low contrast detail (in the grass for example) probably from a noise reducer or multiple re-encodings. It also has poorer deinterlacing (the original sources was 25i, 50 fields per second, and both were converted to 60p with field blending and deinterlacing). Overall CBR is probably better but I would guess that the VBR video looks better at those big bitrate peaks.

How can I check properties of the original source for any video? Is there any way to do that or you say it based on experience?

By examining the samples and based on experience. You can see deinterlacing artifacts on horizontal edges/lines. And you can see field blending artifacts from a frame rate conversion. Watch the star graphic swipe across the screen about 26 seconds into the video. You can see it in other fast moving areas too.

[_Al_]

images CBR vs. VBR, names for these two are swapped! CBR is VBR and VBR is CBR

[_Al_]

some images

[Santuzzu]

Can someone explain the whole process of TV broadcasting? I mean, the whole process starting from acquiring video using cameras to recording it on PC.

[KarMa]

Can someone explain the whole process of TV broadcasting? I mean, the whole process starting from acquiring video using cameras to recording it on PC.

The camera or a nearby capture device outputs a high bitrate video. Then to fit on the internet or a consumer TV broadcast, the video is re-encoded at a much lower bitrate, which makes the file smaller and worse quality.

[Santuzzu]

And who does this re-encoding of the raw video? TV station? What are the usual properties (resolution, bit rate, frame rate, etc.) of the raw video before compression? And how do properties of the TV device affect recording of the broadcast video?

[KarMa]

Could be literally anything. It could be captured on an iPhone (happens mostly with News production) all the way up to a Medium format 4K camera. I have no idea what was used on this random soccer game footage. I know the NFL in the US will capture with 4K cameras, and then use all the resolution as a good form of digital zoom should they need it. Then it gets outputted to 1080i for NBC or CBS, and 720p for ABC and FOX because that's the resolution they broadcast at. Then once that gets to your local TV station, that video will get re-encoded again with H.264 or MPEG2 at a proper bitrate to fit within bandwidth limits. Usually between 6-15Mbits depending on if it's over the air or on cable or satellite.

There are way too many variables to your question and not enough hours in a day. Why exactly are you asking these questions so we can maybe give a better answer.

[Santuzzu]

I don't have any special reason to ask all these questions, just want to learn more about broadcasting. I know I'm asking a lot, hope people will have a patience to answer at least some of these questions.

What about TV devices, how do their properties affect recording? If video shown on TV is 10Mbps, I guess it gets re-encoded once again while it's being recorded. If video shown on TV is 1080i 25 fps, can it be re-encoded to let's say 1080p 50 fps during recording and if yes, who does all this re-encoding at the receiver end? TV device?

[KarMa]

What is a TV Device? Like a DVR? A roku?

When a consumer level TV takes in a 1080i 25 fps video, it will display that video as 1080p 50fps as it deinterlaces the video in realtime using built in hardware. If you record the stream without re-encoding (only an option for video without encryption), then you can record it as the original 1080i 25 fps. It needs to be re-encoded if you want that to be converted to 1080p 50 fps.

[Santuzzu]

And how good is this built-in hardware at the receiver end? I mean, if it takes a soccer match video in a 25 fps, how good would be an interpolation at 60 fps? Is it true that TV stations have better algorithms for motion interpolation?

What if I download a video in a 720p 60fps, how can I know whether it was interpolated by a TV station or by a person who recorded/re-encoded it and uploaded online?

[KarMa]

720p airs at 60fps on US TV.

And how good is this built-in hardware at the receiver end?

It's ok but software based deinterlacing is always better, like QTGMC.

Is it true that TV stations have better algorithms for motion interpolation?

It really depends on who we are talking about. But with de-interlacing, you are doing only in frame interpolation, as an interlaced frame actually has two half resolution frames in one frame. So these two frames are separated and then interpolated vertically to double the vertical pixels, making two full sized frames from one frame. TV stations interlace to save bandwidth while keeping fluid motion of 60fps. Also 25fps makes 50fps not 60fps. 60fps comes from 30fps in the US and other places.

https://en.wikipedia.org/wiki/Interlaced_video

[Santuzzu]

In case of videos I posted here, both TV footages are 720p 60 fps. Since both of them were aired on US TV stations, I suppose they were de-interlaced and interpolated to 60 fps by TV stations since European broadcast standard is 1080i 25 fps.

Now, if I want the best possible quality of recorded soccer matches, does it mean the best for me would be to find or record videos in original quality and then to use software based processing to get desired video quality? What is the best software way to process 1080i 25fps to obtain 1080p 50fps or maybe 1080p 60fps video? Or better to ask, what effective software options I have available to get a quality 60fps video from any source? I have a bunch of soccer matches and not many of them are in 60fps though many of them are in HD. I want the best possible way I can process them and get desired 60fps without many artifacts.

[jagabo]

Honestly I don't think these videos were converted with simple blending because TV stations today use much better converters with motion interpolation. The motion in these videos is good, not many artifacts from conversion.

You are right they are using motion interpolation. Motion interpolation works very well with panning shots and simple motions. It doesn't work well with complex or very large motions. So the algorithm they are using resorts to blending when motions aren't simple. So in wide panning shots the panning is smooth but motions within those shots, and in many close up shots, you can see blending artifacts.

Watch the ball and the feet of the player on the left:

[Attachment 48426 - Click to enlarge]

You can also see that the two videos blend at different frames (because they started at different points in the cycle). So the blending wasn't part of the original 25i video.

CBR on the left, VBR on the right:

[Attachment 48427 - Click to enlarge]

And a few frames later:

[Attachment 48428 - Click to enlarge]

[Santuzzu]

In case of videos I posted here, both TV footages are 720p 60 fps. Since both of them were aired on US TV stations, I suppose they were de-interlaced and interpolated to 60 fps by TV stations since European broadcast standard is 1080i 25 fps.

Now, if I want the best possible quality of recorded soccer matches, does it mean the best for me would be to find or record videos in original quality and then to use software based processing to get desired video quality? What is the best software way to process 1080i 25fps to obtain 1080p 50fps or maybe 1080p 60fps video? Or better to ask, what effective software options I have available to get a quality 60fps video from any source? I have a bunch of soccer matches and not many of them are in 60fps though many of them are in HD. I want the best possible way I can process them and get desired 60fps without many artifacts.

I have to quote myself. What do you think people, if I download a 1080i 25fps original source feed, can I do a better processing than one present in existing footages in 720p 60fps to get a desired 1080p 60fps using the best possible software?

[KarMa]

Probably

[Santuzzu]

Can you give me some recommendation what software are the best for each processing part in case of soccer matches? I have in mind de-interlacing, motion interpolation, and the rest. Processing time is not important, I'm ready to wait as much as it is necessary to get the best possible quality.