Rendering Settings

VideoByDave (vfwTech.com)
Media Composer - Vegas - CS3

Formerly "Avid Liquid" of Colorado

MC4
Tutorials

MC4
Editing

MC4 DVD

Liquid
Vegas
CS3

Compression &
Conversion

Audio

Lights

Cameras

Tech / Legal

Misc

Product Reviews

Liquidians on YOU TUBE Colorado Vidoegraphers ColoradoFilmVideo Assoc Online Video Magazines EventDV () DV.com DV Info.net VideoMaker Video News Avid MC:* AVID MC forum Vegas 8 Editing: VegasVideoHelp.com Squidoo.com-Learning Vegas Ed Troxel's Newsletter Sony - Vegas Pro Forum Creative Cow-Vegas Forum CS3 - Premier Pro Wrigley Premier Pro Site CreativeCow CS3-Premier Pro Audition Soundbooth CS3 Encore CS Flash CS3 Adobe Vid Workshop AdobeTV.com Liquid Editing: Pinnacle AVID LIQUID forum Other Forums DvIndoNet Cameras: (about HDV cams) Canon HV-A1 DVI Forum Canon HV20 Forum Panasonic HMC-150 (AVCHD) Other Creative Cow Sites Business & Marketing Event Videographers Audio Professionals Indie Film & Documentary Cinematography Broadcast Video Check These Out: Learning Center Video University Digital Juice WEVA 4EverGroup

Advanced Rendering Settings

vaman

Posted: Aug 29, 2004 - 16:03

Tony, they are subpixal routines that are only found in some very very high end systems.
They will inevitably improve the quality tenfolds, then the basic default settings,so you have nothing to fear except a bit longer rendering times.
Strom, or premiere doesn’t have them, after fx has very limited implimentation and its also a global implementation.
you can do selective re-rendering if you don’t want to re-render everything.
then change the setting and then render other clips with those settings.
once they are rendered, and you change the settings for another clip, this change doesn’t re-render what was previously rendered with another setting.
on the other hand, remember, that even so if you have to re-render everything, your picture quality will be of much higher quality then storm can ever deliver (or your default already rendered quality)

to test, Try massive zooming in a Dv footage using the old classical 2d or 3d filter.
use these setting set rendermode to YUV processing 16 bit per channel
deinterlacing =5 tap filter
positioning =bilinear
scaling = 6 Lobed Lanczos
you will be stunned.
and then try zooming that much in Photoshop, or after fx, and you will see what I am saying about 10 times quality enhancement

and also remember (if you have to use outside sw, know that the deflicker filter uses, averaging of scanlines, which is not a good choice, as you will loose massive details
, I suggest, you use afterfx, or commotion built-in defliker and storm or premiere one as a last option.

vaman

Posted: Aug 29, 2004 - 12:03

they are all subpixal routines, and affect only the rendered frames.
you have to use it selectively dependent upon situations.
for example sin32 works great for scaling down, etc giving you very high sharpness, with least artifact.

Let me explain what it is and in the end what I find useful in most situations.
Hope you will find all this interesting.

Generally when you apply colorcorrection the value of individual pixels change in a RGB or UV space mathematically, for example (*3,-8)
Generally when a filter calculates individual pixel by examining its neighbors it is called spatial filter. Whereas temporal filters calculate this over a range of frame to update the current pixel accordingly.(used often for slowmotion, variable motion to remove the edgeing often found while doing temporal changes also the other use is often for de-graining, de-interlacing and general video noise reductions),
Both temporal and spatial filter are applied where animated transformation are in use.
Now Edition/Purple is well equipped atleast with Spatial filters which supersamples at various subpixel level.
There is a huge verity available when one is doing geometric transformation like Scale and Rotate.
Although I would expect Pinnacle/fast staff to add some more for colorcorrection and bluring (sharpening, embossing, and median filtering).And also what is sorely missed is routines for temporal or extended de-interlacing and dynamic wrapped slowdowns and speed ups.
Another serious drawback is that one cannot manually (individually) shift colorspace from Yuv to rgb and back for obvious reasons-that this is Dv world and if one is expected to do any real good blue screening in DV then one should have the option to apply slight blur on the UVchannels -a 6 to 10 pixel blur to either U (Cr, or red/green difference) or V (Cb, or blue/green difference) to smoothen the blocky chunks around the keyed edges. Otherwise to do this you need a visualC++ basic interpreter and some good code basics.
Some effects require a separation of fields and a re integration after the effects are calculated.
This feature is sadly misses too
Anyway (you always have X-send)
But when it comes to geometric transformations there already is plenty and probably Edition has the best in the industry.
So It would benefit Edition users to know more as to what happens when Spatial filters are applied for re-sampling an image with (field) that is being geometrically transformed, i.e., after a Scale operation or Rotate.
From My own experience, the best quality is achieved when one uses the SinC filter when scaling down it maintains small details when scaling down with good aliasing but Ringing problems make it a questionable choice for scaling up.
So use Sinc 32x32 for scaling down also one can use 6 Lobed Lanczos which is similar to the sinc, but with less sharpness and ringing.
Lanczos is generally good from most general use (small scaledown and ups) it is well known for its edge-enhancement but some ringing might be still visable..
But for extensive Zoom or scaling I would recommend Cubic B Spline. This is the same high quality Catmull-Rom filter and is also known by the name Mitchel. This is a very balance filter for scaling up for good choice between sharpness and ringing very good edge-enhancement.

There are Others too in the list, and some might work better for static visuals but because of theirs poor handling of high frequencies they often can cause unwanted artifacts.

A wrong use of filters becomes visible when animating a pan as the filter has to deal with increments of less than a pixel (Subpixel). This is often seen as image pop to whole-pixel boundaries as it moves - a visual artifact that is surprisingly noticeable, even when working at high resolutions
So what’s the choice for higher quality renders
For scaling down I suggest SinC
And for scale up as Cubic B Spline
And for general use Lanczos
But remember, these filters are global.