how to deinterlace in most appropriate way?

[baunduley]

i often get confused how to deinterlace in the most appropriate way!
pal to ntsc, ivtc, telecined, the frame rates, the interlaced frame count, purely progressive, hybrid, 3:2 pulldown, 2:3 pulldown, top field, dupe frames, bottom field 2:2:2:2:2:2:3 pull down and all these terms get me totally confused
i've searching all over the internet, and gone through a huge number of guides but it left me confused more

i just want to figure out what how video has been encoded in my retail dvd, and how to deinterlace it in the most appropriate way in avisynth
most of the time i try among 3 to 5 common/popular methods and get decent result
but i really want to absorb the actual logic behind the operation rather than throwing stone in the dark


hundreds of users discussed thousands of time!
if i start another one, will it be a crime??:p

[lansing]

a good place to start:
http://www.doom9.org/ivtc-tut.htm

[baunduley]

a good place to start:
http://www.doom9.org/ivtc-tut.htm

thanx for ur kind help.....gone through this post earlier, couldnt absorb much

can you please explain briefly in simple language? thanx in advance

[johnmeyer]

Why do you want to deinterlace? There are lots of reasons not to do it, most notably that it always irretrievably degrades your video. On the other hand, it is absolutely required when doing certain operations.

You may very well not need to do any deinterlacing, depending on what you are actually trying to do, and therefore may not need to get a graduate degree in video in order to understand interlacing.

The short version is this: video is comprised of rows of pixels. With progressive video, all the rows are captured at the same moment in time; with interlaced video the even rows are captured at one moment in time, and the odd rows are captured at a later moment. The even and odd rows are called "fields." The two fields (even and odd) taken together give you one complete frame of video.

Most of the confusion -- and much of the endless bad thinking and resulting errors -- surrounding interlaced video comes from looking at "freeze frames" of both fields (one frame) taken together. Since each field was captured at a different moment in time, when you bring them together at the same moment in time (which is what a freeze frame does), then the result looks awful. I understand why people do this, but it is almost always the wrong thing to do.

If you want to look at and analyze what is going on with interlaced video, and you want to freeze the video to look at something, then you will get "fooled" less if you first separate the video into fields, and look at one field at a time. This is also the correct way to process interlaced video, although there are many situations where better processing can be done by first deinterlacing and then processing. I don't have time to go into all of that, but when you are scaling interlaced video, you will get better results if you first deinterlace; on the other hand, noise reduction and other similar algorithms can be applied simply by operating on the even fields as though they were one video stream, then doing the same for the odd fields, and then weaving the results back together.