"... as the much lower bitrate from Vimeo certainly acts as a low-pass filter, masking fine detail and noise)..."

Actually, I wondered about that myself.

My HD LCD is not 50" but I do begin to see some objectionable noise and artifacts at about 35%. The shaky footage, one would think, would have to be pretty bad to require zooming in 35%.

In any case, we share the same concerns in this dilemma. I can still see a place for Mercalli in the toolbox, though.

"My HD LCD is not 50" but I do begin to see some objectionable noise and artifacts at about 35%."

Jay,

Does the above statement apply to watching your original clip, or its Vimeo version?

Piotr

when you said "image" I thought you were referring to a still. :)

I've had similar experience when zooming in on HD footage & playing it back in HD. I've figured out the limit depending on the shot setup.

However, the best part is, I can zoom WAY in when playing back as SD. :)

"I've figured out the limit depending on the shot setup."

Would you care to elaborate?

P.S. Not all images are "still". ;o)

"Does the above statement apply to watching your original clip, or its Vimeo version?"

That would be on the original. Sorry for not making that more clear. I did notice that the areas of soft focus were where it did not hold up as well as in the areas of sharp focus.

On the other hand, if I were doing a project for SD or Web delivery (not 1080p), I would be comfortable with 50%, if necessary.

Perhaps the next thing I should try is burning the test to DVD and see how it holds up there.

Your test was most interesting to me.
It highlights something I learned elsewhere regarding why 5D footage can look so sharp when it's well know that it is quite low resolution.
The answer put forward is that it looks sharp because what is in focus is so much sharper than what isn't. Now your test shows exactly this. The in focus lettuce leaves hold up quite well even at 50%, the out of focus leaves pretty well immediately look worse.
I suspect this is also what TheHappyFriar is saying.
It also reminds me of another tip someone gave me some time ago regarding wide shots which can look pretty mushy. Put something in the foreground in sharp focus.

Bob.

Bob & Jay.

Just tested a similar clip, with Mercalli V2 with no border, zooming in as much as necessary + Neat Video denoising & sharpening, in this sequence.

On a shallow DOF scene with the main subject in perfect focus, it can do wonders!

Piotr

"... wide shots which can look pretty mushy."

Yes, I have experienced that myself. South Florida (Miami-Dade County) is really flat--like Kansas with palm trees. When shooting in the Everglades, the wide vistas when shot without something in the foreground look soft, or mushy as you put it. Kind'a defeats the whole purpose of HD.

"... Neat Video denoising & sharpening..."

Great idea! Thanks for sharing. Now I'll have to get Neat Video as well. I understand there's a new version for Vegas.

Can you share your results online?

Our absolutely flat 'outback' has the same problemo, really helps if you can find a dead tree.

Going back to Piotr's musings for a second I feel the issue is fairly complex. In your test that lettuce has a high contrast edge (high MTF) and the body of the leaf has no detail to give the eye a cue. It really highlights how the subject affects perception. When it comes to image stabilisation it gets more complex. If the shot is bouncing around the eye cannot really lock onto the edges, by default the shot will look soft anyway, we don't expect to resolve much detail.
Take the motion away and even without any zooming we may see it as soft relative to how it looked with the motion.

Bob.

Bob, you answered a question I was mulling over yesterday--Why are the stabilized images soft (when I know they are in focus)?

So is Piotr's approach (Neat Video + sharpening) an answer? That's why I asked if it were possible to see some examples.

To sum it up at this stage.

I'm finding motion blur, inevitable with those most jerky, hand-held camera movements, to be the one most limiting factor for stabilized picture quality, including the sharpness. NeatVideo can get rid of the zoomed-in picture's increased perceived noise, and restore apparent sharpness - but obviously it cannot help with those momentary focus losses due to motion blur...

Of course, it's even more visible with zooming-in, as opposed to bordering without zoom - as the objects and their absolute movement distances are getting enlarged.

So Jay - before I post some link to my result example(s), please give me some time to experiment with faster shutter speeds when shooting hand-held.

Piotr

PS That said, I still find the Mercalli V2 trial very promising. One thing though (I asked this in the other thread already): am I the only one seeing memory leak in Vegas projects with Mercalli applied?

If you want a sharper (less soft) image after stabilizing the footage, try this: Add the Vegas Sharpen effect to your stabilized clip(s) and set the Amount to 0.000 (the default setting).

Jøran Toresen

Hi Piotr,

I had strange problem with cooperation between V1 an V2. ProDAD support could not help me, even reproduce problem. I had to deinstall V132, V232 and V264. Then I have cleaned registry (In my case it is TuneUp utilities) and manually deleted all directories conneted with ProDAD in users/appdata etc. Then install once again. Problem solved. Try to do the same to copy with memory issues.

PS. very interesting topic, pair V2+Neat is mandatory in almost each clip in my job, mostly for smoothinig ultra-slow movements on dolly or crane.

Best regards,

Piotr, I look forward to seeing your experiments.

Insofar as the memory leak is concerned, with my limited time on Mercalli V2 I haven't experienced that... yet.

Since I don't expect too much support from ProDAD while I'm only using a demo version of Mercalli V2 - here is another question to those of you guys who have been using stabilizing in Vegas (not necessarily with V2):

- I have already mentioned that one of the factors, limiting successful stabilization, is motion blur associated with jerky camera movements. I've run a couple of tests with 1/100th and 1/150th shutter, and the results are indeed better.

However, I'm still fighting another artifact that before the V2, I used to associate with the rolling shutter skew not handled effectively enough by the previous version (and other plug-ins): sometimes, with a certain combinations of camera's random movements, the stabilized picture gets skewed really bad. In fact, when watching with the border on, one can observe the horizontal and vertical border edges are not at right angle momentarily. It also looks as if the image was moved (get skewed) in 3D, i.e. in the direction normal to the screen....

Assuming the Rolling Shutter correction option in V2 is really working as well as the reviewers are saying, this ugly phenomenon is not caused by rolling shutter - but then, what do you think is the real reason for it? I really can't find a common "feature"in those clips I stabilized with Meraclli V2 so far, that could be suspected responsible for this behaviour... Not knowing the reason, I don't know which direction to tune Mercalli FX so that this "3D / skew" artifact is minimized.

Any suggestions appreciated;

Piotr

"It also looks as if the image was moved (get skewed) in 3D, i.e. in the direction normal to the screen...."

I don't know if we're talking about the same thing, but I've noticed when I apply too much stabilaztion (even in V2), the end result looks like the image is reflecting on a piece of mirror-like mylar film gently moving in a breeze.

When I back off on the amount of stabilaztion, it seems to remove or minimize the effect.

" It also looks as if the image was moved (get skewed) in 3D, i.e. in the direction normal to the screen...."

I've noticed this curious artifact in several cameras, all of which use CMOS sensors. The worst I've noticed it with is the Z5. It looks like one part of the image becomes detached from the others and moves around briefly. Truly wierd and I'm glad someone else is noticing it because so far the only response I've had when I've raised the issue related to my choice of mind altering substances.


I suspect it is a combination of the rolling shutter and the compression scheme or else something else related to how CMOS sensors are scanned beyond what the manufacturers have revealed. I have a suspicion that the sensors are being scanned more than once per frame and the results stiched together. Looking very closely at the motion blur I feel there's something odd about what comes out of some of the latest cameras.

Bob.

OK guys (Jay G. in particular) - after some extensive testing, my conclusion is:

- if you're using a CMOS camera (like the EX1(R)/3), Mercalli V2 is not worth the over $200 price. Their "Rolling Shutter Compensation" is just not effective enough.

This ends my pursuit; no need for posting example results, Jay - sorry :).

Disclaimer: This is not to say other similar plug-in's are any better; it's just that the CMOS artifacts need a lot of more homework to be done by stabilizing software developers in general. ..

Interestingly, and contrary to widely adopted opinions, I never find my EX1 rolling shutter objectionable - apart from when I try to stabilize my footage!

Piotr

Thanks for the information, Piotr. Every littel bit helps!

Jay,

I hope you understood me correctly - I think Mercalli V2 is a great piece of software. However, with full HD in mind as the delivery format, and a rolling shutter camera like the EX1 as the acquisition source - it just doesn't deliver...Not in rolling shutter skew cases.

Let's wait for the Vegas Pro 10 in-built stabilization FX; perhaps it's going to perform similarly in those "simple cases", and it's included in the Vegas Pro upgrade price.

But this is just me - YMMV. Cheers

Piotr

Your conclusion comes as no surprise.
Camera movement creates two artifacts, motion blur and in the case of cameras with a rolling shutter skew.
The first can be removed by using a fast shutter speed but that's the very thing that masks skew. Skew is very complex. For 'simple' scenes such as a flat wall it's pretty easy for software to compute the skew and compensate. For complexe scenes with objects at different distances from the lens moving in different directions it becomes nigh impossible to compute the motion and hence compensate.
One proposed solution was to not only image the scene but to record frames that registered object distance using lasers.
None of this is my ponderings, just a summation of the protracted discussions that went on when the RED camera was first launched.

My own very simplistic view:
If it could be fixed in post then the camera designers would probably put it into the silicon smarts inside the camera.

Bob.