Video Quality compare/meter

[no_alone]

I Wanted to make a video quality meter
the facts i want to comapre
video codec= xvid/divx ...
video source = ts/dvdrip/cam...
video bitrate = 912/1200/... kbp/s
video resulation = 576x624 250x256
video Frames for a sec = 25

My Starting Idea = video bitrate/( resulationX * resulationY * Frame For Sec)

But i need to put the video codec and the source to account... Any idea?

[jagabo]

Try this:

Take an hour long noisy VHS recording and encode it with Divx or Xvid with a constant quantizer of 3. Note the quality, final average bitrate, and file size.

Now take a still frame from that VHS tape and create an hour long still encoded with exactly the same settings. Note the quality, final bitrate, and file size.

See how your quality meter is useless?

[PuzZLeR]

The quantizer tells a true story of quality and I agree with Jagabo. You have to look at things like PSNR, and other properties of physics actually to determine the complexity of video, not what its resolution and other similar metrics are.

Your method is about as useless as a comment saying that all women will fit an exact dress size if they are this weight, or this height, etc. You will never find a matching dress size for all women on these parameters.

Video is just as complex. One video's quality can look great with DivX at 900kbps, while other clips look terrible at 2000kbps.

If you use your system then you will find that a movie of nothing but black screen will be just as good a quality level at 1200kbps as an full-fledged action movie with lots of colors, movement, etc. at the same bitrate of 1200kbps. No way.

Just because they may have the same rez, frame-rate, etc, means absolutely nothing in quality level. Nothing.

[guns1inger]

And then you have to take into account whether or not Paris Hilton stars in it.

So many factors . . . . .

However, if you are in fact serious, have a look at what these guys are doing : http://www.compression.ru/video/quality_measure/index_en.html

[jagabo]

Note that MSU is comparing the quality of the encoded video versus the original source. That is a completely different issue, one that can be objectively measured.

[no_alone]

guns1inger
Yes i sow the website, Thanks , but as jagabo said this is a comparsion tool.
I want to check each file individually .....
I see from your post that this might not be possible...
Maybe someone else can help?

[jagabo]

There is already a tool like you mentioned in GSpot, the Qf factor. Essentially, it's the number of bits per pixel, the same as your above equation.

http://en.wikipedia.org/wiki/Xvid_Qf

[guns1inger]

All this gives you is a number. On it's own it tells you nothing about the actual quality of the video, only how much data has been allocated. A number is worthless when the concept is as subjective as video.

Example : Picking two videos at random from my collection. Both are encoded with Xvid and are 696MB and 699MB in size respectively. They have the following qf values : 0.205 and 0.251

So which is the better quality video ?

[ronnylov]

DivX DRF Analyzer can get some information out of the file.
http://www.free-codecs.com/DivX_DRF_Analyzer_download.htm

I know it does not tell much about actual quality but it can find potential problems with the encoding like if it is compressed to much (too high crf values). Maximum, minimum and average crf values can be interesting. But you can get low crf values by using too aggresive quantize matrices and still have bad quality...

The only correct way is to compare the encoded file with the original file, you need both to see the quality degadation.