proDAD Mercalli workflow with HD clips

megabit wrote on 9/12/2009, 5:58 AM
I have installed the trial of Mercalli deshaker, and have 2 questions to those experienced in using it:

1. From how it works, and considering a number of ProDAD Mercalli avi templates have been added to my system, I assume that the analysis simply renders the file out to avi - however, I cannot find those avi clips nowhere; where are they located?

2. Is it possible to have a 1080p clip deshaken without scaling (with static black borders), and then simply rendered out to say 720p, or even PAL Widescreen - so that the border is cropped away, and the center portion of the original 1920x1080 frame (stabilized) becomes my new, lower resolution, but not upscaled frame? What I'm seeing now is that whatever my Vegas project / render out settings, the Mercalli frame need to be scaled up in order to get rid of the border - which is not what I need. I'd imagine that creating a border within the full 1920x1080 frame would allow me to crop an SD frame from its stabilized center and render out to PAL, but for some reason I cannot achieve my goal...

What am I missing?

TIA,

Piotr

AMD TR 2990WX CPU | MSI X399 CARBON AC | 64GB RAM@XMP2933  | 2x RTX 2080Ti GPU | 4x 3TB WD Black RAID0 media drive | 3x 1TB NVMe RAID0 cache drive | SSD SATA system drive | AX1600i PSU | Decklink 12G Extreme | Samsung UHD reference monitor (calibrated)

Comments

megabit wrote on 9/13/2009, 7:09 AM
Guys - I'm sure somebody knows where the ProDAD-Mercalli.avi files are stored. It's funny, but I cannot find it - and my stabilized clip plays back at full quality / full speed, so it must be playing from some prerendered file!

AMD TR 2990WX CPU | MSI X399 CARBON AC | 64GB RAM@XMP2933  | 2x RTX 2080Ti GPU | 4x 3TB WD Black RAID0 media drive | 3x 1TB NVMe RAID0 cache drive | SSD SATA system drive | AX1600i PSU | Decklink 12G Extreme | Samsung UHD reference monitor (calibrated)

Jøran Toresen wrote on 9/13/2009, 7:36 AM
Use Pan/Crop to get rid of the static black borders. Then render your clips.

Jøran Toresen
megabit wrote on 9/13/2009, 7:57 AM
Since I'm interested in specific output resolutions only (like wide SD PAL, or better still, 720p) - I'm using using another approach: rather than pan/crop within a full HD project, I simply change the project properties to my output resolution and render out. This way I can get the center cropped away from the full HD frame, and without the border, too.

But I'll repeat my question: where do the bloody Mercalli renders go on my HDD?!

AMD TR 2990WX CPU | MSI X399 CARBON AC | 64GB RAM@XMP2933  | 2x RTX 2080Ti GPU | 4x 3TB WD Black RAID0 media drive | 3x 1TB NVMe RAID0 cache drive | SSD SATA system drive | AX1600i PSU | Decklink 12G Extreme | Samsung UHD reference monitor (calibrated)

R0cky wrote on 9/13/2009, 8:34 AM
Mercalli doesn't render until you render your project. I stores an analysis file somewhere, wherever other plugins store their settings data probably.
srode wrote on 9/13/2009, 11:14 AM
I believe it stores the analysis file in the .Veg file - I compared before and after adding 1 Mercalli event analysis and the .veg file went from 121 to 268kb, nothing else was changed in the project.
megabit wrote on 9/13/2009, 1:11 PM
Yep, I can confirm the veg file is rather large...
If this is they way it works, then I must admit it's quite efficient - as I said, the stabilized clip plays back at full speed/quality just as it did before the anaylysis!

One thing is not so good about Mercalli, and I'm afraid it's a show-stopper to me: the end result is sort of "jelly", and not natural looking at all. Has anyone found the settings' "sweet spot" for their application (each case is different, of course) - or is it a common knowledge ProDAD Mercalli is not able to output a more natural results?

AMD TR 2990WX CPU | MSI X399 CARBON AC | 64GB RAM@XMP2933  | 2x RTX 2080Ti GPU | 4x 3TB WD Black RAID0 media drive | 3x 1TB NVMe RAID0 cache drive | SSD SATA system drive | AX1600i PSU | Decklink 12G Extreme | Samsung UHD reference monitor (calibrated)

farss wrote on 9/13/2009, 3:45 PM
"is it a common knowledge ProDAD Mercalli is not able to output a more natural results?"

Image stabilisation will only remove the movement of the camera, it cannot correct what the camera recorded in the frame.
Two problems arise.
Firstly the motion blur from a normal shutter speed will still track the original camera movement. Never found anything that could fix that in post. Solution I have used is to set shutter speed quite high the one time I shot something I knew would need deshaking in post.

Second problem you might have is skew from a rolling shutter. Only footage I've shot that I stabilised was from a Z1 so no problem with rolling shutter. The Foundry do have a plugin for AE that can correct the skew but I doubt it could fix skew in multiple directions causing the jello effect.

Bob.
megabit wrote on 9/14/2009, 1:41 AM
Bob,

I must admit that watching the output, the rolling shutter effect did cross my mind. However, I don't quite understand why it is so pronounced after the Mercalli analysis (it is virtually invisible in the original footage).

Piotr

AMD TR 2990WX CPU | MSI X399 CARBON AC | 64GB RAM@XMP2933  | 2x RTX 2080Ti GPU | 4x 3TB WD Black RAID0 media drive | 3x 1TB NVMe RAID0 cache drive | SSD SATA system drive | AX1600i PSU | Decklink 12G Extreme | Samsung UHD reference monitor (calibrated)

srode wrote on 9/14/2009, 3:41 AM
Have you looked at individual frames in the origional footage?
farss wrote on 9/14/2009, 3:58 AM
Agree with srode. Take a long hard look at it frame by frame. Turn on the grid in the preview window. You only need a tiny amount of jello in a stable image to make it look really obvious.

Bob.
megabit wrote on 9/14/2009, 4:42 AM
Well, I did as advised; here are 2 jpegs taken from two consecutive frames of my source material (sorry for the quality - it's not important). If you put them on Vegas t/l and play back as frames, on transition between them, the bottom of the pole moves much more than its top. But I'm still uncertain whether it's the rolling shutter at its worst, or simply that my panning started with a little rocking of the camera...

Frame1

Frame2

AMD TR 2990WX CPU | MSI X399 CARBON AC | 64GB RAM@XMP2933  | 2x RTX 2080Ti GPU | 4x 3TB WD Black RAID0 media drive | 3x 1TB NVMe RAID0 cache drive | SSD SATA system drive | AX1600i PSU | Decklink 12G Extreme | Samsung UHD reference monitor (calibrated)

farss wrote on 9/14/2009, 4:47 AM
Maybe this forum's problems are worse than we thought but I'm not seeing any images or links, sorry.

Bob.
megabit wrote on 9/14/2009, 4:50 AM
Yep, images don't work - I provided links instead.

AMD TR 2990WX CPU | MSI X399 CARBON AC | 64GB RAM@XMP2933  | 2x RTX 2080Ti GPU | 4x 3TB WD Black RAID0 media drive | 3x 1TB NVMe RAID0 cache drive | SSD SATA system drive | AX1600i PSU | Decklink 12G Extreme | Samsung UHD reference monitor (calibrated)

megabit wrote on 9/14/2009, 4:55 AM
sorry - now the links should work

Piotr

AMD TR 2990WX CPU | MSI X399 CARBON AC | 64GB RAM@XMP2933  | 2x RTX 2080Ti GPU | 4x 3TB WD Black RAID0 media drive | 3x 1TB NVMe RAID0 cache drive | SSD SATA system drive | AX1600i PSU | Decklink 12G Extreme | Samsung UHD reference monitor (calibrated)

farss wrote on 9/14/2009, 5:37 AM
Go them.
Tried quickly to align the two frames using PS, no joy but I didn't persist.
Made a 2 frame sequence in Vegas and looped it.
Not nice to look at. Really hard to say for sure if there's any jello going on there or not. Certain the frame is undergoing rotation. I think that is something that image stabilisers have a hard time of, especially if it's happening very quickly. There's a lot of research being done by Adobe, was a topic here a few months ago. Their work showed quite dramatically how image stabilisation could create some really wierd problems with bendy buildings and the like.

Bob.
megabit wrote on 9/14/2009, 5:44 AM
Not nice to look at. Really hard to say for sure if there's any jello going on there or not. Certain the frame is undergoing rotation

Exactly my thoughts, too.
Of course, you normally cut out and trash places like this, but I used Mercalli on it on purpose. The effect is extremely jelly.

AMD TR 2990WX CPU | MSI X399 CARBON AC | 64GB RAM@XMP2933  | 2x RTX 2080Ti GPU | 4x 3TB WD Black RAID0 media drive | 3x 1TB NVMe RAID0 cache drive | SSD SATA system drive | AX1600i PSU | Decklink 12G Extreme | Samsung UHD reference monitor (calibrated)

R0cky wrote on 9/14/2009, 7:02 AM
I see weird effects that could be described at jelly or other image distortion or just jumps in the output when the applied correction is too aggressive. This happens both with Mercalli/Vegas or deshaker/Vdub.

I have found that the presets in Mercalli are pretty good for the situations described in their names - they give about the maximum stabilization you can get without getting these other types of image problems.
TheHappyFriar wrote on 9/14/2009, 9:26 AM
looks liek the camera was focused on the fence/tree in front. It would be pretty hard to use an out of focus object as the stable object, it's already blurry. The fence/tree look very sharp compared to the "rocket".