Hybrid source problem

[rasiina]

I have a hybrid DVD that has been badly deinterlaced. Every time camera angle changes the top field of next frame contains crap from previous frame before camera angle change. So i need a way to find all these frames (or if not possible i can manually select the frames) and after that interpolate the full frame from bottom field using help from next frame or just resizing it. Any suggestions how to do this?

[pc_speak]

Hi rasiina
Sounds like it is interlaced still with the Bottom Field First (BFF) or maybe TFF. You have ripped the video to your HDD?
Use DGIndex to create an AVS script file as input to virtuladub and try via this method.
http://www.gromkov.com/faq/videoedit/deinterlace.html and use the internal filter 'Deinterlace-->Duplicate Field 1' or maybe 2. This might remove the bad fields throughout the whole movie. Then again the movie may look worse. :D

[Leak]

use the internal filter 'Deinterlace-->Duplicate Field 1' or maybe 2. This might remove the bad fields throughout the whole movie. Then again the movie may look worse. :D
Errr... why not do the same in AviSynth directly?
SeparateFields()
SelectEven() # or SelectOdd()
LanczosResize(last.width,last.height*2)
And believe me, the movie *will* look worse if you do that on all frames.

I wonder if TFM couldn't be used to detect and "fix" those scenechanges - could you perhaps post a clip with one or two such scenechanges in it?

[Guest]

Both FieldDeinterlace() and TDeint() can be run in full=false mode, where only combed frames are deinterlaced. You may consider using one of them to postprocess your video, assuming of course that something else is not happening that calls for a different treatment. As Leak suggests, a sample source video would remove all doubts about that.

[rasiina]

Yea, sample is better than bad explanations.

http://www.sharebigfile.com/file/170681/sample-vob.html

Just first place i found to hold files in google. You can get the file after a small wait, hehe.

Most of the movie is that kind of video, but there are parts that are correctly interlaced too. I can easily deinterlace those parts.

I used tdeint(full=false) in my experiments.