MKVToolNix mkvmerge shrinks video bitrate by 7%

[draig-llofrudd]

I muxed from M2TS 1080p to MKV. The M2TS file has got 12719 kBit/s audio stream. Now after changing the container via muxing with mkvmerge, the video bitrate is now decreased by 11887 (about 7%)! I thought that shouldn't be the case as changing the container doesn't encode the video, or at least a vastly smaller difference of like up to 1%. So why is this?

Okay, there may already be some threads explaining this problem and referring to "overhead". I'm not familiar with some terms. But in my video I can even see a little difference. There occur horizontal violet lines all over the video when frames are switching because of rapid camera movemend. These lines may be visible very very weak and only be noticed when concentrating on them and immediately having the eyes in front of the monitor, but they are still present. Should this be? Is this normal? I mean it doesn't bother me at all, but I am still interested in "fixing" it if possible.

[sneaker]

How did you measure the video bitrate?

Possible reasons:
1. mkvmerge removes filler NALUs
2. mkv container overhead is much smaller than transport stream overhead
3. you did not measure correctly
4. a mix of 1, 2, 3

[draig-llofrudd]

This is where I've looked it up. And sorry, the M2TS bitrate is a little more than 12719. My bad. Is this how I can compare it or must I do a calculation? And even if, why can I see a difference in the video? By the way, this doesn't affect all muxes, but mostly the 1080p videos.

[KarMa]

Mediainfo will give you more info on your files.

[sneaker]

Oh, you were talking about overall bitrate, not specifically about video track bitrate. It's ok how you compared. Just looking at the file size should yield the same result without relying on any software doing deeper analyses.

And even if, why can I see a difference in the video?

You are probably imaging things or it's just a problem with your player. If you are still unsure you could compare them more systematically, e.g. decode and compare checksums of raw decoded data or compare in AviSynth. (http://avisynth.nl/index.php/Compare or http://avisynth.nl/index.php/Subtract for example)
But IMHO that will be a waste of your time. There will be no difference.

[poisondeathray]

The decoded stream will be identical. Just demux the elementary video stream and compare (e.g. tsmuxer for ts or mkvextract for mkv). That is the proof. MKVMerge will remove things like sequence headers, so there might be a few bytes difference, but the actual decoded video stream will be the same. If you want to triple check, you can do a per frame psnr check, that is definitive proof.

It might "look" different depending on how the player is configured. You mentioned "1080p", but some players are configured to deinterlace transport streams

[draig-llofrudd]

As sneaker already mentioned, I would waste my time comparing the videos unnecessarily since it doesn't bother me, because the difference is practically zero. But I will leave a screenshot in a few minutes just to show you what I mean.

Edit: Well, it only appears in motion of the pictures, so I can't make a screenshot. It's quite distinct. But let's leave it by that. Here's what it roughly looks like (did it with GIMP):