to deinterlace or not to deinterlace...?

[balkandvb]

I have a finished project in Final Cut Pro that is regular interlaced 29.97 fps video and now before i make it a SVCD Id like to know if someone has any comments on wether i should deinterlace the video in FCP then rip it to svcd or if I should not deinterlace.

I tried to not deinterlace to svcd and the video didn't APPEAR interlaced, the movement as a matter of fact looked deinterlace. Is this a requirement for SVCD?

[edDV]

If the goal is to play it on a DVD player to a TV, do not deinterlace. The TV will expect an interlaced SVCD.

If the goal is PC only playback or computer streaming, then you shouldn't be using SVCD at all and you may want to deinterlace your MPeg depending on the computer player targeted.

[balkandvb]

thank you for replying

actually, the goal is to play on a DVD player, BUT, also make it not look like the "home video" feel and more like the film feel.. or as close to 24p as possible.

[lordsmurf]

http://www.digitalfaq.com/capture/interlace.htm

[balkandvb]

Good link smurf, but..

Im not sure if you guys are familiar with FCP but the following is a way that I make the dv progressive and get the motion blur i want and i think it is a good way to without loosing quality:

video track 2: de-interlace filter, even fields, opacity 50 percent
video track 1: de-interlace filter, odd fields

The resulting video does not look like home video anymore and it is progressive with a little motion blur and it has really no deinterlace artifacts. I make no frame rate changes.

[edDV]

thank you for replying

actually, the goal is to play on a DVD player, BUT, also make it not look like the "home video" feel and more like the film feel.. or as close to 24p as possible.

Well CDs are cheap. First make an interlaced SVCD as your "reference".

Then play with variations and experiments to get that film look and compare what you get to the reference. It's a good way to learn.

To really see what is going on, you need a good progressive DVD player hooked up to a good progressive TV (EDTV or HDTV). The DVD player will convert interlace 29.97 frames per second to progressive 60 frames per second for display. If it detects a 3:2 film source, it will reconstruct the 24 frame per second sequence and then display that at 60 frames per second by repeating frames in a pattern that adds up to 60.

[balkandvb]

alright well my dvd player is progressive but my tv is not.. just standard 480i..

But I will see any differences with the Final cut deinterlace method and no method.. thanks.

[edDV]

http://www.digitalfaq.com/capture/interlace.htm

Nice guide Lordsmurf. Somehow I missed that one before.

The only expansion I can offer is in the area of adaptive de-interlace. Note that is refers to realtime deinterlace not MPeg encoding errors. Crude deinterlace offers two choices.

1. Toss one of the fields and repeat the remaining field twice.
This gets you a poor man's progressive image at 29.97 fps but suffers from loss of 50% of vertical resolution resulting in the stair steps on diagonals that you demonstrated.

2. Merge the fields using the "bob" technique
Best to temporal interpolate the missing lines from fields before and after in time. This seems like a great idea until something moves in the image. All motion areas go blurry.

Video engineers long ago figured a solution based on human vision characteristics. The human eye detects less detail in moving objects but is sensitive to temporal blur. So, the solution is to measure motion areas in the image by comparison with previous and subsequent frames (using both fields for fancy algorithms) and use this motion information to switch between #1 and #2 above on a pixel by pixel basis. This results in de-interlaced frame detail for low motion areas and half vertical resolution but temporally sharp (no motion blur) pixels for motion areas.

This technique is used more often than you think in live TV. Any time an image is scaled (zoomed), rotated, filtered or somehow placed into 3D space, it often needs to be deinterlaced prior to digital filtering to maintain a quality image. Of course, after the effect is generated, the output is re-interlaced into 2D TV space for the picture you see.

The same concept would apply to generating a "film look" that will be played back on a normal TV.

[lordsmurf]

I need to read that again when I'm more awake.
Bookmarking.

[edDV]

Good link smurf, but..

Im not sure if you guys are familiar with FCP but the following is a way that I make the dv progressive and get the motion blur i want and i think it is a good way to without loosing quality:

video track 2: de-interlace filter, even fields, opacity 50 percent
video track 1: de-interlace filter, odd fields

The resulting video does not look like home video anymore and it is progressive with a little motion blur and it has really no deinterlace artifacts. I make no frame rate changes.

Hmm,

I'm not sure exactly what the FCF deinterlace filter is doing, but assuming a perfect deinterlace, what would reducing opacity of even fields do?

First, it would darken the image ~25%. This could be compensated with another filter.

Second, it would partially reduce vertical resolution putting luminance 100% to 50% steps every other line. This is not very filmlike.

Third, it would introduce some jerkyness to the image reducing the effects of 60 fields per second half way to 30 fields per second. Even if you play single field video at 30 fields per second the result is more like flicker rather than the 24fps motion stuttered look of film. Note that film projectors typically repeat frames twice for ~48fps flicker so that may be their target look.

Seems to me that they are just introducing flicker, not displaying 24fps motion.

Other factors in "film look"

- different luma and chroma gamma profiles. Filters exist for this.
- film color saturation. Filters exist for this.
- moving film grain (noise). Filters exist for this.
- dust and scratches - Filters exist for this.
- picture shake due to perf and film gate tolerance. Filters exist for this.
- 18-24fps motion - this could be done by true frame interpolation
- film dynamic range - hard to match (DV is compressed into 8bits)

[balkandvb]

Well, Im not aware of what the opacity does, but I read this method somewhere and believe me, it looks f**ing great on a STV... Im aware of the other "film look" factors but I was concentrating on the motion characteristics.

Remember that video track 2 is a duplicate of video track 1, and that the result is 30p, not 24p, but it has just eliminated the too smooth video look with no NOTICEABLE image degredation.

[balkandvb]

Here is the link to where i found it. Take a look at "deinterlace blend"

http://www.kenstone.net/fcp_homepage/feature_video_film.html

Notice the author rates that method as the best method to achieve the look.

[edDV]

Well, Im not aware of what the opacity does, but I read this method somewhere and believe me, it looks f**ing great on a STV... Im aware of the other "film look" factors but I was concentrating on the motion characteristics.

Opacity is transparency. So I think what you are describing is taking the two 720x240 fields, placing the first over the second at 50% transparency and then expanding the frame back out to 720x480. If they output that at 30fps progressive then they have "blended" fields as described in lordsmurf's link. That is not filmlike.

Remember that video track 2 is a duplicate of video track 1, and that the result is 30p, not 24p, but it has just eliminated the too smooth video look with no NOTICEABLE image degredation.

Before you said the two layers were the even and odd fields.

Play some motion or scene changes in single frame steps on your computer or camcorder and see if you see double images. Also see it it looks different on the computer monitor (progressive) vs. TV (interlaced)

Somehow they have to blend field motion that is displaced 1/60 sec into one frame to to get 30fps progressive.

The other way to it is as I described above with motion compensation.
Do you have a link to this effect?

[balkandvb]

So wait, im confused.

who are you trying to prove wrong here.. me or the author of the link i posted above of the FCP method?

[edDV]

Sorry, I didn't see your post while I was writing mine. I'll read it now.

[edDV]

RE: http://www.kenstone.net/fcp_homepage/feature_video_film.html

We are into "art in the eye of the beholder" realm here. The last three apps seem to have various handles for dealing with deinterlacing two fields which are offset in motion by 1/60 sec.

Mr. Coplan never addresses motion in his analysis other than "motion blur". Film motion blur results from the shutter being open while the background and/or forground is in motion relative to the camera. For example, the camera pans with a forground actor walking. The actor can remain in focus while the background exhibits motion blur due to greater relative motion, regrardless of being in optical focus or not. The effect can vary with different f stops, shutter speeds, lighting and lenses.

Now if you do this with Mr. Coplan's "Deinterlace Blend" (blending two fields offset in time 1/60 sec) in progressive display, the forground actor would exhibit a double image with one image brighter than the other based on the relative opacity of even vs. odd fields. The background would show a wider offset double image due to greater relative motion. This may or may not be an interesting effect but it is not "motion blur" in my book. Show me a technique that keeps the forground actor spacially entact and the background in a blur not double vision. If the image is stationary, then Mr. Coplan's observations would hold. He needs to put those test charts into motion.

Joe's Field Blender seems to be just that, a weighted blur of the two fields but in this case it calculates a blur of fields not just a mix. To me this still low tech.
http://www.joesfilters.com/joesfieldblender.php

ReelSmart Motion Blur - does it with pixel motion tracking between fields. A much better approach
http://www.revisionfx.com/mblur.htm

DVFilm Maker also uses motion analysis to deinterlace or deinterlace + interpolate to 24p
http://www.dvfilm.com/maker/index.htm
From their FAQ,
"...Nearly all plug-in and built-in de-interlacers reduce vertical resolution by 50%. But DVFilm Maker works by analyzing motion in each frame, and only changing pixels in areas of the screen in which there is movement. This results in a much sharper image especially for scenes where the camera is still."

DigiEffects Cinemotion (plug-in) TBD nothing found



Maybe my analysis is wrong too. Anyone else?

[balkandvb]

That is great info.. but you are digging too deep. All I wanted was the non video look.

Have you ever seen the Panasonic DVX-100a's 24p mode? I have, and it looks awesome. People have even put 16:9 anamorphic converters on there and converted the dv 24p/16:9 footage to film. Now of course, in no way does it replace a film camera, which has much much better contrast ratios than dv cameras, but its good for the "MTV" feel.

I tried to carefully compare joe's field blender from to the "deinterlace blend" method and there is hardly a difference. Coplan even says they do the same thing.

And what about the second method he mentioned

Video 2 track: opacity 30
Video 1 track: lower deinterlace

Will this achieve the "background in blur but not person effect?"

I will look into the other things you mentioned.

[edDV]

I like the other two in concept better but if we are talking about artistic effects, the best is in the eye of the beholder.

The Panasonic DVX-100a is a great camera. The 24p mode is perfect for Indi Cinema film transfers or for "film look" progressive DVDs. Beyond 24p this camera has gamma filters to enhance the film look.

http://www.adamwilt.com/24p/

[lordsmurf]

So wait, im confused.
who are you trying to prove wrong here.. me or the author of the link i posted above of the FCP method?

From what I can tell, the author in the link likes ghosting artifacts.
Easy as that.