"true resolution" of video [Archive]

Slavko_1

20th May 2007, 15:41

how can I compute a real/true resolution of video material ? Many video conversions are upsize to some standards like DVD, SVCD from lower resolutions. Are there any mehods to know about "true" resolution of source material ?

joseph5

20th May 2007, 17:52

GSpot or MediaInfo.

Slavko_1

20th May 2007, 18:27

Sorry but this programs gives only "nominal" resolution as width x height. It is not information about "really" resolution. I can resize from 1 x 1 to 1920 x 1080 without any troubles but "really" resolution is still 1x1. But how compute it having only just upsized material ?

Blue_MiSfit

20th May 2007, 18:37

Well, by your definition of "really" resolution, it sounds like you want to know the resolution of the actual source.

For example, a DVD backup with 640x480 resolution that came from a 720x480 DVD, that was actually transferred from a 480x480 SVCD that was made from a LaserDisc or something like that :)

My advice: Don't worry about it.

Even knowing what the actual original resolution of a source is doesn't really change how you would process it. Just do the best you can with the source. Who cares where it came from? :)

~MiSfit

CAFxX

20th May 2007, 18:55

Perform a DFT on one or more frames. If you take what its output is with a grain of salt you should be able to find that information.
This kind of analysis is normally performed on the outputs of cameras and camcorders to measure the "real" resolution of the entire system (sensor, demosaicing, compression, etc) and not just of the sensor.

Wilbert

20th May 2007, 19:57

how can I compute a real/true resolution of video material ? Many video conversions are upsize to some standards like DVD, SVCD from lower resolutions. Are there any mehods to know about "true" resolution of source material ?
That's not always possible. Details and thus resolution is lost when downsizing.

You can try Image Analyzer (a VDub filter):

http://trevlac.us/Analyzer.zip

It more or less does what CAFxX suggested (i think).

Slavko_1

21st May 2007, 10:05

Sorry but:
How to perform a DFT ? Many thanks for script or something ...

yup

21st May 2007, 12:47

That's not always possible. Details and thus resolution is lost when downsizing.

You can try Image Analyzer (a VDub filter):

http://trevlac.us/Analyzer.zip

It more or less does what CAFxX suggested (i think).
Wilbert?
Where can find RFFT2Wdll.dll for Analyzer?
yup.

CAFxX

21st May 2007, 13:07

Sorry but:
How to perform a DFT ? Many thanks for script or something ...

Matlab, or the like.

*.mp4 guy

21st May 2007, 17:25

There is a plugin for the Gimp that can do forward and inverse ffts aswell.

wonkey_monkey

21st May 2007, 17:37

If it's been upsized, you can try downsizing to different sizes (perhaps in one dimension only to make it easier). At some point the image might seem to "snap" into place (a lot of the stairstepping from the upscale will disappear) and you can be pretty sure that's the original resolution.

David

Slavko_1

22nd May 2007, 09:12

Thanks davidhorman. I made this:

a = AVISource ("Test.avi") # testing material 720x576
b = LanczosResize (a, 704, 560). LanczosResize (720, 576)
compare (b, a, "", "compare.log")

I will change resolution step -16 and when "Mean Absolute Difference" dramatically grow, one step earlier will probably my "true resolution". Is this good idea ? (sorry my bad english)
One question. What will if video was not only upsized but also some noise was added (for example), sharpness .... ?

Wilbert

22nd May 2007, 20:55

Wilbert?
Where can find RFFT2Wdll.dll for Analyzer?
yup.
See attachment.

ronnylov

23rd May 2007, 13:05

Thanks davidhorman. I made this:

a = AVISource ("Test.avi") # testing material 720x576
b = LanczosResize (a, 704, 560). LanczosResize (720, 576)
compare (b, a, "", "compare.log")

I will change resolution step -16 and when "Mean Absolute Difference" dramatically grow, one step earlier will probably my "true resolution". Is this good idea ? (sorry my bad english)
One question. What will if video was not only upsized but also some noise was added (for example), sharpness .... ?

Seems like a good idea to try but as davidhorman said do only one dimension first and then do the same thing with the other dimension. But the "true" dimension may differ from different scenes of the original file. Perhaps bad focus of the camera or motion blur in certain scenes lowers the "true" resolution. I think you should look foor a scene with lot of small details.
When they test the "true resolution" of a camera they normally use a resolution test chart but I guess your source does not include that.