helge79
21st January 2005, 05:42
I am trying to rescue as much information as possible out of an improperly telecined NTSC DVD :confused:
If I separate the fields with
AssumeTFF().SeparateFields()
and examine the output in VirtualDub, I see the following telecine pattern:
Imagine there has been a sequence of frames in the original 24fps movie
ABCDE...
or, in terms of the corresponding fields,
(At Ab) (Bt Bb) (Ct Cb) (Dt Db) (Et Eb)...
Looking at the fields of the successive DVD frames, I can identify them as follows:
At
Ab
Bt blended with At
Bb
Bt blended with Ct
Cb
Ct
Db blended with Cb
Dt
Db blended with Eb
and so on...
In other words, there is a 3-2 telecine pattern, but it is implemented in a malevolent way. I shall not discuss what should be done to the people who've made the DVD (following the Pulp Fiction, http://www.doom9.org/ivtc-tut.htm suggests "No trial, no jury -- just straight to execution." :devil: ) What I am trying to understand is what's the best way to deal with the situation at hand.
In principle, no information is lost in the above telecine process. For example, one could subtract a fraction of At from the blend of Bt and At and rescale luminocity to restore the original Bt. In practice, of course, the resulting Bt turns out very noisy because of the discreteness of the color space and the compression artifacts.
So, the only reasonable solution that occurs to me is to restore the A-frame from At and Ab, and to double the vertical resolution in Bb to create a new frame B', i.e. the resulting video is
(At Ab)
(__ Bb)
(Ct Cb)
(__ Db)
...
The 5-to-4 inverse telecine is performed properly and the ghosting is avoided, but every other frame suffers a 1/2 loss in vertical resolution. (A possible improvement is to retain the information from the static (no-ghosting) parts of the blended fields -- even though I am not quite sure how to implement it in AviSynth...)
Does anybody have any comments/suggestions regarding this procedure? Also, if anybody understands how the blended frames could occur during the original telecine process, it would also contribute to my general education (even though it is not relevant for performing the inverse telecine).
Thanks! :D
If I separate the fields with
AssumeTFF().SeparateFields()
and examine the output in VirtualDub, I see the following telecine pattern:
Imagine there has been a sequence of frames in the original 24fps movie
ABCDE...
or, in terms of the corresponding fields,
(At Ab) (Bt Bb) (Ct Cb) (Dt Db) (Et Eb)...
Looking at the fields of the successive DVD frames, I can identify them as follows:
At
Ab
Bt blended with At
Bb
Bt blended with Ct
Cb
Ct
Db blended with Cb
Dt
Db blended with Eb
and so on...
In other words, there is a 3-2 telecine pattern, but it is implemented in a malevolent way. I shall not discuss what should be done to the people who've made the DVD (following the Pulp Fiction, http://www.doom9.org/ivtc-tut.htm suggests "No trial, no jury -- just straight to execution." :devil: ) What I am trying to understand is what's the best way to deal with the situation at hand.
In principle, no information is lost in the above telecine process. For example, one could subtract a fraction of At from the blend of Bt and At and rescale luminocity to restore the original Bt. In practice, of course, the resulting Bt turns out very noisy because of the discreteness of the color space and the compression artifacts.
So, the only reasonable solution that occurs to me is to restore the A-frame from At and Ab, and to double the vertical resolution in Bb to create a new frame B', i.e. the resulting video is
(At Ab)
(__ Bb)
(Ct Cb)
(__ Db)
...
The 5-to-4 inverse telecine is performed properly and the ghosting is avoided, but every other frame suffers a 1/2 loss in vertical resolution. (A possible improvement is to retain the information from the static (no-ghosting) parts of the blended fields -- even though I am not quite sure how to implement it in AviSynth...)
Does anybody have any comments/suggestions regarding this procedure? Also, if anybody understands how the blended frames could occur during the original telecine process, it would also contribute to my general education (even though it is not relevant for performing the inverse telecine).
Thanks! :D