Laurence
Are you saying that you (we) can’t stabilize 16:9 video using DeShaker in Vertualdub?

Joran

You can Deshake 16:9 video, but the problem is that there is no extra picture to fill in the edges.

To understand what I am saying picture how Deshaker works. You run through the video clip twice, the first time you log all the shaking movement. The second time it compensates for each shake in the footage by shaking the entire frame the opposite way. Without edge compensation, this would mean that what you would see is the outside edge moving around erratically with a stable image in the middle.

There are two ways to deal with this outside edge: you can zoom the image and crop or you can substitute bits of the last frame that filled in that edge. Zooming after Deshaking means that you are going to lose resolution, so most of us prefer the latter approach. Deshaker is unique in that it can use that it can fill in these edges. The problem with filling in the frame edges is that especially on the top and bottom of the frame, the old frames don't exactly line up. Let's say that a person is walking across the frame and you are applying Deshaker to the clip. What you'll see are the bottoms of his legs dubbed in offset from where they should be. It looks pretty bad. The same thing happens with the tops of peoples heads.

On a regular 4:3 TV you can get away with a lot of this because it happens in the safe zone and you really don't see it. With a 16:9 image however this is not the case. 16:9 TVs are digital LCD or Plasma designs and seem to have a smaller safe area. Even worse, when you look at the 16:9 image on a regular 4:3 TV, the top and bottom of the frame are in clear sight. If you use Deshaker on 16:9 footage in this context what you will see are offset leg bottoms and head tops regularly.

This is where 4:3 to 16:9 cropping and aspect ratio changes is so nice. If you shoot 4:3, deshake the footage in this format, then render it to 16:9, almost all of this cutting and pasting of the frame edges happens in an area that you are cutting off anyway. In the final render you see nice solid top and bottom edges, but in reality the image at these edges was compiled from the extra area from the 4:3 frame. You can still see the cutting and pasting happening at the right and left edges of the frame, but that is a whole lot subtler. On pretty much all TVs there is still a little safe area there and if it goes beyond that, at least you don't have shifting body parts of the main characters!

What I'm finding is that letterboxing Deshaken 4:3 footage looks really good and gets rid of most of the Deshaker edge problems. Rendering this letterboxed 16:9 image to 16:9 anamorphic widescreen may not be as high resolution as a native 16:9 image, but being able to Deshake it without the top and bottom edge artifacts means that in some cases, the final product in some ways can actually be better.

Laurence

Thank you for a thorough explanation. But, can you give a more in depth description of the content in the last paragraph? How and where do you do this? (I’m not English, so another question: Does letterboxing mean black borders on top / bottom of the video?).

Joran
Norway

Well if you are in Norway, you are in a PAL format country.

In either NTSC or PAL, there are two different aspect ratios that video is presented in: standard screen and widescreen. For a long time there was just standard screen, but in order to better present widescreen movies, a new widescreen format was developed. A chief requirement of this new format would be that it would be 100% compatible with the older squarer format. The way they did that was this:

A new format was introduced: widescreen. This new format could be played back on either a widescreen TV or an older 4:3 TV. If it was played back on a widescreen TV, the same number of pixels would be shown, but each of these pixels was 25% wider and so the overall image was 25% wider as well. If it was played back on a regular TV, every 4th line was dropped and thus a correctly formatted 16:9 image would be displayed on the 4:3 television.

Since this was mostly done to improve the vertical resolution of movies transferred from film, video cameras remained 4:3 native and everyone was happy.

Then two things happened: One, people started using DV format for small theatrical releases and putting black bars on the top and bottom of the screen to simulate the widescreen look. Two, widescreen TVs became more common. The big problem here is that if you put letterboxed 4:3 footage on a 16:9 screen, instead of the image filling the screen, you would get black bars around all four sides and a small image in the center of the screen.

Then cameras like the PD150 and others started adding "widescreen" modes. They did this by cropping the image and doubling or interpolating every third line. This would format okay on a widescreen TV, but did this at the price of vertical resolution.

At this point in time the situation is this: Standard screen is an archaic format. While many CRT 4:3 televisions are still sold and in use, they are gradually being fazed out in favour of 16:9 widescreen sets. 16:9 sets can show 4:3 footage just fine, but it requires the television to be manually set this way and most viewers either don't know how or don't bother to do this.

Vegas lets you change from 4:3 to 16:9 aspect ratios, but doing this doesn't give you the extra vertical resolution you would have if you had a fancy new 16:9 native camera. Basically all that Vegas does is crop the tops and bottoms of the 4:3 frame, and interpolate the extra lines needed to bring the stretched image back to 720 x 480 (or in PAL land 756 x 480) pixels. Many of us here have been lusting after new cameras that would do a full resolution 16:9 image.

What I was pointing out in my initial post was that if the extra pixels at the top and bottom of the screen are quite usefull if you want to use VirtualDub and Deshaker on your raw footage and give you results you simply couldn't get with footage actually shot in 16:9 format. Also that, on computer playback or on playback on a regular 4:3 TV, that extra resolution is thrown away anyway, and especially then the end result of Deshaken and letterboxed 4:3 footage is actually a little better.

I haven't had to use Deshaker with anything I've shot in 16x9 with my XL2 but: If you go to Pan/Crop on a clip and zoom in on your image ( keeping the box in Pan/Crop the same ratio ) you will basically have the same thing that the XL2 is doing with 4:3 but you'll still have the option of keeping the increased resolution of 16x9 on those shots that didn't need Deshaker. Essentially all the XL2 is doing is cropping the edges so in fact you do have a little wiggle room if you need it.

- Ray
- Ray

www.undergroundplanet.com

Laurence

Thanks again. So what you do is this:

1. You stabilize 4:3 video using Deshaker in Virtualdub (without cropping the video), and save the footage in 4:3 format (using for example the MainConcept DV codec).

2. You start a 16:9 project in Vegas.

3. You crop the stabilized 4:3 video (with “moving” edges) using the 16:9 preset in the Event pan/crop tool in Vegas or by right-clicking and choosing Match output aspect.

4. You can combine these clips with (“real”) 16:9 video (which has the pixel aspect ration 1.4568 here in PAL-land).

5. You export your finished video in 16:9 format.

Am I correct in my interpretation of what you are doing?

Joran

Guys,
I gotta jump in here ...

I think I understand the long posts by Laurence, but ... here's my personal experience with DeShaker.

I shot footage, lots of it, on the 16:9 setting of my VX2000 (NTSC). I very successfully used DeShaker and never had a problem with edge correction. There is a setting in DeShaker that lets you set how many frames forward and backward it looks to extrapolate the edges after deshaking (Use previous and future frames... setting).

There is also a setting in DeShaker where you specify the aspect ratios.

With my footage, for NTSC, I set both the Source pixel and Destination pixel aspect ratios to - Anamorphic NTSC (1.215)

I couldn't be happier with the results of DeShaking when viewing the resulting DVD on a widescreen TV.

Tom

Except for the "combining with real 16:9 footage", yes that's it. My VX2000 only shoots 4:3, so all my 16:9 footage is cropped 4:3. I automate the aspect ratio with Ultimate-S, but that is doing the same thing. Of course, animated photos and titles are rendered directly into the 16:9 format.

Here is how this applied to me recently.

My wife has a non profit organization which delivers backpacks filled with school supplies to poor South American Children. Normally the children and I travel with her and I shoot this event. This time I stayed at home with the children while she went, taking my NTSC VX2000 with her. She came back from the trip with a bunch of footage that her uncle had shot with my camera that was so shakey I originally thought it was unusable. I did a little research and found out about VirtualDub and Deshaker. I ended up with a 24p 16:9 final project that looks very good, especially considering the shakey handheld nature of the original footage. In this case, the end result was much better than if it had been shot on a 24p 16:9 camera. If I'd had shakey 16:9 footage, I wouldn't have been able to rescue it to this degree. That is what amazes me so much.

Hey Laurence - gotcha.

But you said My VX2000 only shoots 4:3. Is that because you choose so, or does your VX200 somehow not have a 16:9 setting?

Tom