 The
"Preferences" panel to the right is taken from the current version of HDLink.
In the RED circle are shown the 5 quality selectors
available when recording CineForm Intermediate files. Equivalent settings
are also available when recording or rendering from NLE or compositing
applications.So the question is - "what
setting should I use, and what will the resulting bit rate be"?
First some background. CineForm files are
coded using a VBR (variable bit rate) temporal Wavelet algorithm. This
means as image statistics change (image motion, edges, texture, noise profile,
etc) the bit rate goes up and down to compensate. Variable bit rate
algorithms are also known as a constant-quality algorithms.
Based on the VBR nature of the algorithm, the
resulting compressed file size will vary, for any given compression quality
setting, based on the complexity of the image being coded. Obviously different
frame sizes (spatial resolution) and frame rates also change data rates and file
sizes.
As shown in the table there are 5 quality
settings provided: Low, Medium, High, Film Scan, and Film Scan 2. On the
Mac, CineForm quality levels are (currently) represented by a slider and
percentage as shown in the table below. Also shown in the table are approximate
data rates and file sizes for a given quality setting assuming source material
is 10-bit 1920x1080 24p YUV.
Some Comments About the Table:
- 1280x720 material will be about 60% the size
of 1920x1080 material for a given frame rate because of its smaller frame
size
- CineForm 444 material will be about double
the size of CineForm YUV material for a given frame size
- Certain applications don't exploit the
temporal nature of the CineForm algorithm, and resulting YUV file sizes are
approximately 25% larger than shown in the table. This doesn't impact
visual quality, only compressed file size. These applications include
Sony Vegas and Apple Final Cut Pro.
Recommendations:
- Bottom Line: For most workhorse
acquisition, post, and rendering uses we recommend either High or
Film Scan. By the way, the background on the naming of "Film Scan"
is that this mode was designed to accurately reproduce the characteristic of
film grain during film scanning. So which one should you use?
Experiment yourself - see if you can tell the difference.
By the way, our first feature film "Dust to Glory" used the High setting
throughout post (we didn't have the FS modes at that time). The film
prints for D2G were created directly from the "High" quality CineForm
content on the Adobe Premiere Pro timeline and looked great!
- What about Film Scan 2? Internally we
call FS2 "Overkill" - we don't use it very often ourselves. But it's
there for a reason - you may want to consider using FS2 when you have very
demanding post workflow requirements in which you are pushing or stretching
the images a lot - a real lot. Otherwise FS1 and High are the best
settings.
- What about Medium? For modest
post manipulation of images Medium is fine and the image sizes are smaller.
- And Low? For those that need smaller
file sizes Low will work fine, although we generally don't recommend Low for
content that will go through many generations of rendering.
When we performed the
visual fidelity comparison of CineForm 444 versus the respected HDCam SR
format we used both FS1 and FS2. As you can see from the diagrams, both
quality selections exceeded the visual quality of HDCam SR.
Also For consideration: When selecting a
higher quality setting (FS1 versus High for instance) the recorded files are
larger. Be aware that files recorded using higher quality settings demand
more CPU for real-time playback than the same file recorded at a lower quality
setting. In case you care why...this is because larger files sizes require
more entropy decoding (hence more CPU) than the same content recorded with a
lower quality setting.
|
