Is there a program for analyzing video files?

[redeemer-dk]

Is there a program that can load, for example, an xvid video file and let you look at how it is compressed? Where you can see which blocks of the image has been reused from the previous frame and which quantizier has been given to a frame, and what the frame type is? That would be lovely.

Thankyou.

-Rune Svendsen

[sh0dan]

Gspot (nice name ;) can give you an overall image of the AVI file.
ffshow can show the motion vectors used.

Otherwise I can't give you any hints on how to view what's inside a the files. If you are a coder you can modify xvid or ffdshow to give you the information you need.


(og velkommen til :)

[redeemer-dk]

Thanks :) Do you know a place to download those programs? Especially the ffshow one.

(tak for velkomsten :))

[Ookami]

http://www.doom9.org under "Software" :)

[redeemer-dk]

is it ffshow or ffdshow? if it's ffdshow how do you use it then, how can i see the motion vectors? also, when you say "software" you mean "downloads" right?

[sh0dan]

ffdshow can be downloaded from this homepage (http://sourceforge.net/projects/ffdshow/).

gspot is in the doom9 software section.

ffdshow discussions are in the xvid forum!

[cult]

ffdshow/configuration->tick show motion vectors

[HurlyBurly]

I have written an Xvid Quantizer Parser that will work for Koepi and Nic stable builds (no B-Frame detection yet).

The Xvid Quantizer parser can:
-Parse Quantization values of I-Frames and P-Frames for Xvid AVI;
-Get statistics such as maximum, minimum and average quant value;
-Get textual/graphical histogram distribution of quantizers
-Get graphical time-based distribution of quants for I/P-Frames
-Do Ranged analysis (ignore end credits for exemple)
-Save/Load reports of analysis and share with friends!

Binary and source can be found on sourceforge.net (see post below)

[majerle]

@HurlyBurly
Sound very interesting.
but i can't see the attach ( don't know whay, mod is busy ?)
Please post a link or mail it in PM

Thanks

Andres

[sh0dan]

Very interesting - I'm also hoping for the source. This could be an interesting filter for AviSynth (overlaying the information onto the video) - or if it could be possible using a standalone player.

If I understand correctly different MB's can have different quantizers - is this currently used in Xvid? If so - overlaying the quantizer onto the MB could also be interesting.

(it seems like I'm in brainstorm mode, so I'll just continue ;) )

Also - would it be possible to display which pixels are predicted from the ME and which pixels are corrected.

I guess this could bring invalueable data to the XviD developement for optimizing quality.

[HurlyBurly]

Sorry about that attach file problem.

Edit:
The project is now at xvidqp.sourceforge.net (http://xvidqp.sourceforge.net).


Please note that the source is still messy ;)

Also, i forgot to update the readme which says it only work with Koepis 04102002 build but has since been tested and proven to work smoothly with Nic's Sept 9th build too...

[DaveEL]

Originally posted by majerle

but i can't see the attach ( don't know whay, mod is busy ?)


Sorry was away over the weekend and forgot to check the new attatchments this morning.

DaveEL

[majerle]

@DaveEL
No probs at all :p

Andres

[MoonWalker]

Hi,

I tried Xvid Quantizer Parser but I run into some problems.

I loaded up the Avi, pressed analyze and the prog crashed after the bar had gone to the end. I found a analysis.log in the folder but I don't think the quantizers are correct...

14:31:38 - 23 PF:Q27
14:31:38 - 24 PF:Q27
14:31:38 - 25 PF:Q27
14:31:38 - 26 PF:Q250
14:31:38 - 27 PF:Q27
14:31:38 - 28 PF:Q27

MoonWalker

[majerle]

I have the same problem too. I logged 1500 frames, encoded in cbr 1000 (no bf or mpeg tools).

17:02:20 - 1 IF:Q4
17:02:20 - 2 PF:Q31
17:02:20 - 3 PF:Q31
17:02:20 - 4 PF:Q31
17:02:20 - 5 PF:Q31

Sounds not very well.


Andres
P.S i attached the full log file.

[HurlyBurly]

Doesn't look very well indeed. Specially MoonWalker with quants over 31x... hehehe. Would it be possible to know which build you have both tried it with?

The parser has been tested with AVIs encoded with 2 builds:
-Koepi's 04102002 Stable build
-Nic's September 9th 2002 build

And it does hang for a few seconds when analysing large files (but it'll pull through ;) ), have to look into that as well...

[MrBunny]

I don't know what could be wrong with the Parser for everyone else (or so it seems), but I'm using Koepi's Oct 4th Stable Build and the Parser seems to be working flawlessly. Comparing output stats from debugview with the results of the Parser are coming out exactly right. Just my 2 cents to try to ensure people don't get completely scared off :D
Thanks for the great tool.

Mr. B

[^^-+I4004+-^^]

HurlyBurly,would it be possible to make
some markings (on vertical/quant axis)
so we have some indication what quant was where...
now i have many "redlines" and it would be pretty
hard to do some statistical analysis to see
what quant appeared how many times (and that way see which is which) etc.
(ie. horiz. axis is time (is ok already) and vertical should
have marks:2,4,6,8,10.... so we can see what quant was used on particular
frame/time.....)

hope you understand what i mean..
(divx drf analyzer has this scale so it's easy to see
what DRF was used on what point in time... it also has framesizes graph,but i didn't said that.......(heheh))

cheers_

[HurlyBurly]

I understand what you mean perfectly. I will add it for sure :) In the meantime, here is some info that could be useful that I didn't mention anywhere.

In the time-based view:
Light green = Q2
Dark green = Q3-Q5
Yellow = Q6-Q9
Red = Q10+
Blue = I-Frame (All Q)
Thanks a lot for the feedback.

[^^-+I4004+-^^]

>In the time-based view:
Light green = Q2
Dark green = Q3-Q5
Yellow = Q6-Q9
Red = Q10+
Blue = I-Frame (All Q)


nice!
thanks and keep the good work...

[Dali Lama]

@HurlyBurly

Great job! I have always wanted a tool like this for Xvid ever since I used DivxAnalyzer for getting the quants used in an encode. Note, I havn't tried it yet, but I hope it works well.

Oh and those are some nifty ideas Sh0dan.

Good Luck,

Dali

[Arska]

I wonder why it's not working for me. For my encodes it reports drf between 0-255. I have Koepi's Oct-4 build installed. I even tried uninstalling and reinstalling the codec. I used Avisynth 2.07 and XP SP1a to encode. I also had xvidqp working just fine with a previous XP installation.

[HurlyBurly]

@Arska
I really couldn't tell you whats wrong since i haven't even tried XP yet hehe. The tool has been successfully tested on almost all other Windows platforms (minus NT4). Sounds oddly like an unsupported build symptom but you say it isn't. All I can say is I hope it works again for you. Try and pet it ;)

@Everyone
Released Beta3 of XvidQP which fixes three small memory leaks, has a speed improvement for analysis process and plots an additional graph (frame sizes). No support for recent Xvid builds yet (hopefully soon if i feel like debugging). Enjoy.

[nfj]

My latest version of DivX DRF Analyzer suports DivX3, DivX4, DivX5 and XVID, I,P,B and S frames (exception for B frames in DivX5 wich are incorrectly acounted for statistical information).

It also doesn't require any codec to be installed.

http://www.geocities.com/analyzerDRF


NFJ

[Valky]

Originally posted by nfj
My latest version of DivX DRF Analyzer suports DivX3, DivX4, DivX5 and XVID, I,P,B and S frames (exception for B frames in DivX5 wich are incorrectly acounted for statistical information).

It also doesn't require any codec to be installed.

http://www.geocities.com/analyzerDRF


NFJ


Yes, this is awesome program! I wish I could have found this earlier than last week :)