> Guide Index

Removing Interlacing

If you happened to edit your video interlaced (and I really hope you didn't, but alas, this shall remain here for completeness sake), you will definitely want to remove this interlacing now. If you removed interlacing before editing, then you should obviously skip this step. Interlacing looks terrible on progressive displays like computer monitors and it makes good looking compression difficult.

There are many ways to remove interlacing. However, as your video has been edited together, some of these options are less successful than they would be before you started editing. So let's look at some possible techniques.

OK, so we just
IVTC, right?

Usually the best way to remove interlacing in anime is to use Inverse Telecine (IVTC). This matches up the fields in order to restore the progressive frames. These fields are usually in a pattern, so finding and combining them is usually not that difficult. However, your video has been edited. This makes things very tough indeed. Imagine one pattern of telecine being faded into another pattern of telecine... that pattern now becomes very difficult to find. Footage that is sped up will also have completely lost some fields, making it even harder to find the pattern.

IVTC also includes decimation. If you've done lots of tight editing, beat matching, flashes and all sort of other video trickery, it may be ruined by the decimation process as the footage is reduced from 29.97fps to 23.976fps.

So you can see now the basic problem that we have in removing interlacing with edited videos. Add to this the numerous problems that you can find in sources (hybrids, field blending etc etc) and it becomes a real mess.

So, what can you do?

There are basically two solutions to the problem. You can attempt to perform IVTC on the footage, or you can deinterlace it. I strongly recommend that you try both methods and see which one works best for your video.

Before you begin: Get the correct colorspace

It is likely that at this point you have an avi that has been exported losslessly from your editing program - probably in an RGB colorspace. For pretty much all the filters on this page you will need to convert this to YV12 (the exceptions to this rule will say what you need to convert it to instead). To do this conversion make sure your avisynth script looks something like this:

1) Try to IVTC anyway (might work pretty well)

It may be that your video is edited together in such a simplistic way that IVTC will be just as good now as it would have been working on the original source. It's not very likely, but it's possible. If you are going to use this method, a basic script like this will do:

If you have PAL footage, set decimate=0. If you are using NTSC, you can still try using decimate=0 and see if you like the result better. [Note: for more information on the AMVIVTC function, see the page on making your footage progressive]

2) Full Frame Deinterlace (most likely solution)

Not always the prettiest option but certainly one of the more reliable techniques. There are many different ways to deinterlace things, and so you don't have to remember how to use a ton of different deinterlacing filters, I have created the AMVDeint() function which will allow you to very easily try different methods. AMVDeint have a parameter called mode (NOT related to the mode parameter in AMVIVTC!) which can be set from 0 to 4. Each mode specifies a different type of deinterlacing, with the lower values representing faster methods, and larger values representing the slower methods. Try them all out on your source to see what you think works best for your situation.

One of the downsides of using the AMVDeint function is that you cant tweak any settings on the deinterlacers if you want to. I think you normally shouldnt need to tweak any settings on them, but in case you do, here is a list of the deinterlacers that are used, in case you want to call them normally.

AMVDeint(mode=0) - This mode uses the LeakKernelDeint filter, which is a very fast deinterlacer with alright quality.

AMVDeint(mode=1) - This mode uses the Sangnom filter, which tends to give particularly good results on anime.

AMVDeint(mode=2) - This mode uses the Tdeint filter, which is a very good deinterlacer which interpolates fields.

AMVDeint(mode=3) - This mode uses the Tdeint filter with the settings TDeint(mode=2, mtnmode=3, blim=100). This makes it blend fields together which some people find more pleasing than interpolating.

AMVDeint(mode=4) - Ok, now we start getting more complicated. This mode calls the Tdeint filter, and uses the TMM filter to create a mask of the interlaced parts of the images, and finally uses the NNEDI filter to deinterlace those areas. This tends to look very nice but is also very slow.

Once you have a script with deinterlacing, you will want to check out how to crop and resize your video.