Sharpening: in-camera vs. in post

megabit wrote on 8/28/2010, 11:57 AM

As I wrote in another thread, I'm a happy user of the Neat Video plug-in for Vegas. I find that shooting with detail enhancement off in the camera, and then de-noising and sharpening to taste using NV brings great results.

However, somewhere at the back of my head doubts still remain whether this really is the optimal workflow... It has some obvious advantages:

- no need to decide on whether use sharpening or not on the camera
- by eliminating sharpening artifacts, the encoder (EX1 and/or the nanoFlash in my case) can be used even more effectively.

On the other hand, who said Vegas encoding (with or without Neat Video sharpening - one can also use the Unsharp Mask or Sharpen FX's for this) will always perform a better job that in-camera detail enhancement?

So as you can see, my Neat Video plugin workflow question has been extrapolated to some more general dilemma:

Most people say it's better to shoot with detail enhancement off, and only sharpen in post. Is this opinion well grounded? Does it hold true with all NLEs / codex / sharpening FX's?

Comments welcome!

Piotr

Back to post

Comments

MarkWWW wrote on 9/2/2010, 11:35 AM

When a similar question about the Convolution Kernel arose a few months ago I posted this explanation - perhaps it may help to explain some things for you too.

Mark

farss wrote on 9/2/2010, 2:24 PM

"Bob, you saw an example of my video denoised and sharpened with Neat Video - what do you say?"

In a technical sense the results are fantastic, in a subjective sense I felt they were for some reason that I cannot easily explain; 'plastic'.

I've used Mike Crash'e free DNR plug to do much the thing, mainly because I've found getting rid of all noise gave me much better outcomes on footage uploaded to Youtube. Maybe they have since improved their encoders / scalers, Youtube is a moving target.
At one time I was finding my original 720p uploades looked fantastic but the downscaled versions looked liked shite. Denoising the footage cured that.
On the other hand I've also found not doing any de-interlacing before downscaling which looses some vertical resolution i.e. my 720p ends up as 1280x540 actual res gives much the same outcome, curious. The advantage with this approach is it's fast to render and requires no careful tweaking.

Bob.

GlennChan wrote on 9/5/2010, 9:22 AM

For sharpening, you don't really need to know about or use the convolution kernel (this is my opinion). The unsharp mask can be more powerful if you need a larger radius. The sharpen FX is simpler to use if you want a basic sharpen (and sharpen FX is probably faster than unsharp mask).