"I see it as what you are stuck with if you don't like what comes out of your camera"

If we're talking about the "tactics", then I'd agree - however, I was meaning the "strategy"...

:)

Piotr

Typically HD cameras use considerably less in-camera sharpening anyway, but my preference is doing the sharpening in post if it's necessary. Your workflow seems perfectly viable to me.

I almost never use the "Unsharp Mask" plugin and only occasionally use the "Sharpening" filter, instead I use the "Convolution Kernel" with some custom settings. I came up with some detail enhancement and focus enhancement settings for the "Convolution Kernel" filter that work really, really well. I'm not currently at a computer and I don't remember the settings off the top of my head so I'll have to post them later. (Remind me if I don't.)

To expand a bit, I use the same tactic as when recording audio -- I want everything I can get on the tape and worry about any enhancements in post.

Example 1: If I apply equalization in the audio recording stage, I may compromise frequency bands I will wish I had later. In post I have a choice.

Example 2: If I apply sharpening during the shoot, I run the risk of noise in the blacks or oversharpening the edges. In post I have a choice, and am not stuck with a possibly bad choice I made earlier.

Since I don't do live broadcasts, I'll take the control over the uncertainty every time.

....and this is exactly the strategy (not tactics) I'm using, musicvid.

If I bring the matter forward to discussion is because I'd like to make sure the advantages to it (those I mentioned in my OP, and those you mention above) are not outweighed by e.g. a certain combination of encoding formats / NLE used / sharpening filters bringing actually worse results, than detail enhancement in this particular camera hardware would.

And if so, what such a combination would be, and under which circumstances....

Just an academic discussion I was hoping for, to expand my own and other participants knowledge / experience. I hope I'm clear now.

Unfortunately the analogy between what you'd do in the audio and video worlds doesn't hold up so well unless you're using a camera that's recording RAW. To make the comparison a bit more valid I'd change your perspective to consider what you'd do on the audio side if you were recording to mp3 and at a fairly low bit rate.

The camera, well the better ones, do the calcs on the 14bit data from the sensor's A/D converters and then crunch all of that down into a quite lossy codec as 8 bit data with low chroma sampling. I'm not saying that means you should do everything in camera just that it makes this discussion rather difficult to get to a conclusive answer.

In my view getting rid of noise as much as possible in the camera, before compression should have obvious advantages over trying to wrangle it in post and indeed too much Detail in camera can lead to serious issues down the track especially if shooting progressive which I generally don't.

To date the best technical quality images I've got from my EX1 shooting something on a stage came out of the last show I shot. The venue had no stage lighting, just enough to meet safety standards so I had to supply my own. So I had 1x 400W HMI spot and 2x 150W HMI fresnels. The stage did not have black blackout cloth, it was very deep blue, maybe they got it cheap :)

Needless to say the shadows on the stage were hideous so in a creative sense it looked really bad. However between the the two 250W parcans that came with the venue and my HMI lights the camera settled on a CT of 4,800K and I had enough light not to have to use gain, yeah! I wound Detail down, Coring Up a long way, black gamma and low ket sat down. I used STD 3 gamma.

Once I got that into Vegas and downcoverted it to SD PAL I was pretty amazed at how good it looked i.e. how clean it was. I did try applying some Unsharpen Mask AFTER downconversion by nesting the HD project however render times were getting really silly so although it did look better, this was a freebie and they wanted the DVDs in a hurry so I left that out.


On the general question of noise in video I had a bit of a wakeup call a couple of weeks ago. We took the movie I've been working on the soundtrack of to a decent preview theatre. The only vision I have is from DVCAM dubs from the HDCAM out of the old F900 and on my monitor it looks like shite in the night time exteriors. The noise in the shadows looks hideous, great big globs of it. This preview theatre had a big screen and a big 2K projector so I was really fearing the worst for how it would look. I was quite taken aback at how friggin good it looked. This was only SD DV and quite amazingly it did not look all that bad at all and the noise did not look anywhere as bad on that big screen. Sadly my mix was another matter entirely but that's another story.

Bob.

No lack of understanding what you are asking, Piotr, it's just that at my age my sense of caution is usually pre-emptive, whether one calls that strategic or tactical (it's probably just reflexive).

But since you've got the equipment to do it, and I don't, I'd be fascinated to see some comparative tests, using motion, static detail, and combinations of the two. I've uploaded a bunch of rez charts that may be useful.

http://shell.dim.com/~musicvid/Test

However between the the two 250W parcans that came with the venue

That's got to take the prize for a low-rent house. Bet they didn't have their BMI certificate, either. . .

"I'll have to post them later. (Remind me if I don't.)"

Consider yourself reminded... I'm very interested to learn about alternatives to Sharp and Unsharp Mask in Vegas!

Hi,

We produce video content as a part of marketing. There is no documentary or reported production at all. All scenes are artificial and carefully staged. Terget medium is salesman's laptop, rarely internet site. We have prepared some compendium based on customer demands (not general, Pareto rule). Most important in compendium is :

- to keep quantity of plans low (max 3) in order to achieve "sterile" filling of picture
- to replace details with big, solid surfaces enrichted with light gradients in order to achieve "CGClike" look
- to avoid natural light and to build scenes with extremely wide scope of light (some parts of scene are quite dark, some are carefully exposed).

Working with such assumptions, I can see that sharpening is not a clue. Clue is the NOISE (this is very similar attitude like in camera movement : strights path is not a clue - clue is smooth acceleration and deacceleration or constant, very low speed).

Using both : sharpening and noise reduction in camera AND Neat Video denoiser in POST, we can organize all parts of scene. I take care to polish first plan in camera (because final output is small, I can use sharpness "in plus" and coring "in plus").

In POST I compose the scene on two tracks. Neat is implemented on lower track (mostly noise reduction simoultanously with sharpening) and has influence on backgrounds. Masks with gradients are implemented on upper track around first plan areas. My target is to keep first plan without post production touch (very natural, rather video then movie look, detailed) and to polish backgrounds in order to achieve glosy, plastic, artificial look.

Composition of such prepared scenes and 3D computer generated content or texts looks very coherent.

Best regards,

AFAIK, sharpening is not a reversible process.

It can also screw up certain post production tasks like chroma keying and (to a lesser degree) secondary color correction.

The argument FOR sharpening in camera would be:
A- Workflow. No rendering if you don't plan on adding your own sharpening/NR/whatever. Sometimes rendering is slooooooow.
B- The sharpening algorithm in your NLE is not as good as the camera's. (You probably won't notice this though.)

" The sharpening algorithm in your NLE is not as good as the camera's. (You probably won't notice this though.) "

If you can explain how to achieve the same level of detail control in Vegas that cameras such as the EX1 provides I think a few of us would be very interested. I guess one could nut out a way to do it but it would be fairly tedious plus a huge render hog compared to just the basic Unsharpen Mask.

Bob.

Bob? - If you can explain how to achieve the same level of detail control in Vegas that cameras such as the EX1 provides I think a few of us would be very interested.

Haven't you just made Glenn's point for him? . . Confused-of-London . .

- g

I wasn't disagreeing with him but what if you do notice?

There's good reasons not to apply sharpening in camera. As Glenn rightly points out you cannot undo it and if you have done it you can face a nightmare in post if scaling down to SD especially if it's interlaced.
So my point was to try to get some better insight into how to do really good sharpening in post because doing it well in post can also be problematic and you may notice that you're making noise worse.
The CineAlta cameras offer a considerable array of 'detail' adjustments but trying to translate them into Vegas is confusing, hard enough to understand what they do in the camera and the translation is made worse because camera speak and post speak is different, heck camera speak isn't even consistant between cameras.

One FX that fascinates me in Vegas is the Convolution Kernel but it's not a simple beast. I thought more information on how to use it could be beneficial. I can Google that term just like anyone else however (again) translating what one learns from a Wiki into the actual tool in Vegas is not so simple.

Bob.

John Cline said,

"I almost never use the "Unsharp Mask" plugin and only occasionally use the "Sharpening" filter, instead I use the "Convolution Kernel. (...)"

I've been getting some useful offline feedback on a recent video I made from JohnMeyer and the Convolution Kernel was also mentioned. I need to sharpen things yet tend to over-contrast and over-sharpen on one level and am trying to learn new and better ways.

I've been using the Sony Sharpening so far, I've never used the Kernel. I Googled around to get some tip/tutorial on what it does, etc., found nothing of use. The closest show-and-tell was the non-narrated video at YouTube shown below--yet I don't know if it refers to the Vegas one.

So...

Q-1:

Q-2:

Again, no need to go into complex mathematical algorithms which will go over my head.
Just something a couple of nano-quarks above "For Dummies" explanations :)

Thanks.

Below: Convolution Kernel Video (or so it seems)

Bob and others
With a similar question to the original. but with audio, is my theory right or wrong. When I shoot a talking head. as my camera is only 16bit would I be better to compress and eq while laying to tape. or perform those functions in post. Ive never tried it while shooting. My theory is the pre manipulated 16 bit signal might be cleaner than in post which takes the 16bit manipulaes it and throws away what it doesnt want. so I ask you as you understand these matters of rocket science better than I.
Thanks in Advance
Steven

If you've got HUGE variations in levels causing your quietest sounds to be recorded at very low levels AND you need to add gain to them in post then and only then I'd say some analog taming of the signal would be in order. Something as simple as a good limiter to tame things like doors slamming could be quite useful.

Another preprocessing option that is useful is the low cut filters in many microphones. If you've got wind or rumble using the low cut filter can give you more head room to record with whilst not clipping the unwanted LF noise.

Reason I'm cautious to recommend using any preprocessing is it is night impossible to undo and 16bit audio has a quite significant dynamic range, way beyond what video has. Of course if you've got noisy preamps in your camera then adding some gain between the mic and the camera may help or not as happens with the old PD150.

Bob.

Could someone at some point reply to my Qs on Convolution Kernel and Sharp/Unsharp two postings above this one? Merci beacoups.

Here are three Convolution Kernel settings for Enhance Focus, Enhance Detail and Enhance Edge:

Thanks John,
that's very helpfull. After I looked at those I did some Googling and found this laymans explaination of the Convolution Matrix from the Gimp people. It thankfully avoids all the mathematics, I can almost understand it. Maybe it's enough info to inspire others to have a go at trying their own matrix values.

Bob.

I have messed with that convolution kernal many times and I still don't know what the hell I'm doing with it. Maybe if they made it "automatic for dummies".
JJK

Thanks John for your settings - but are you sure of the last one? Instead of Enhancing Edges, it acts like the "Find Edges" preset...

Piotr

Oops Sorry - I got the middle matrix value wrong (8 instead of 9)...

Yeah... There certainly is a great potential for picture manipulation using the Convolution Kernel FX. However, I think its best use is some special effects - what Neat Video offers is much more straight-forward, intuitive yet powerful:

- separate Y, Cr and Cb sharpening (with the "Conservative" security option)
- separate sliders for the High, Middle, and Low detail frequency channels.

I also hope that - as the sharpening is done in one go with noise reduction - the results of the two are mutually optimized.

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

Piotr