|
Tutorials
QuickTime Video Compression for CrystalMaker
|
CrystalMaker provides industry-leading digital video
capabilities, allowing you to record video as you work,
build, edit and save animations - and export rotatable
QTVR objects. The technology used to do this is called
QuickTime®.
This article provides some general tips for making the most
of QuickTime's powerful video compression settings, and
includes the results of our own tests on video output quality
and file size.
|
|
|
Video Compression
Many people are familiar with the notion of compression,
in the context of making computer images smaller. Common
image compression formats include the ubiqutous "JPEG"
files, found on most websites.
With video, the problem of file sizes is much worse than
with static images. Videos need to display many images per
second, so the file sizes quickly mount up. Developers have
come up with a wide range of algorithms for compressing these
video sequences - some algorithms are just like those used
for compressing still images; others are more advanced, and
compare a sequence of frames, so that only the changes between
frames need to be saved.
Controlling QuickTime
Video compression button
|
|
CrystalMaker's Video Palette (Window > Palette > Video)
includes a Compression Settings button (with an "accordion" icon).
Pressing this brings up a dialog that's crammed with controls.
CrystalMaker is giving you full access to the extraordinary
wealth of control settings provided by QuickTime. These
include video frame rate, output size/quality controls -
and a choice of many different video compressors, also
known as "codecs".
The question is: which codec should
you use - and what particular settings are going to
work best when recording your video?
|
|
CrystalMaker's Compression Settings dialog.
The compression type ("codec") is displayed in the
popup menu at the top of the dialog and the output
quality can be previewed in the bottom right-hand corner
|
|
Criteria for Choosing the Best Video Settings
The key factors you need to consider when recording
video are: image quality and file size. Unfortunately, these
two factors tend to oppose each other, so that the best
image quality tends to result in the worst (i.e., highest)
file sizes. However, some codecs are more efficient than
others, so is possible to make some general recommendations.
By default, CrystalMaker uses the Animation compressor.
This is a good, safe choice. It is designed to work well with images that contain lots of
sharp edges: the type of "synthetic" images used in
programs such as CrystalMaker. The downside is that this
compressor isn't particularly efficient.
By contrast, more efficient video compressors - such as
the popular H.264 codec - don't work very well with
CrystalMaker's images. These type of compressors are designed
for natural images - movies of real objects, where there
are softer transitions, and where things don't change very
much from frame to frame. If you try to using this type of
compressor with "hard" synthetic objects, then you'll end
up with fairly poor results.
Video Testing
To help you choose a good, general purpose codec, we
undertook a detailed series of tests, using a "typical"
CrystalMaker file, and comparing the file sizes and
video quality for the different compressors/settings
used.
Test Settings.We used the Fullerene ("bucky ball") structure, displayed
as a ball-and-stick model, viewed
in red/blue colour stereo, in a graphics pane with dimensions
800 × 600 pixels and using millions of colours.
For each test, the structure was rotated through 60 frames
(at 6° intervals), making a complete revolution. We used
the highest-quality rendering mode (Rendering > Full Rendering Mode),
to give smooth bonds.
|
|
Results of video tests, showing the
output file size and our (subjective) quality ranking.
|
|
| Codec |
Setting |
Size
[MB] |
Quality(*)
Ranking |
Comment |
| Animation |
Millions / Best |
29.7 |
1 |
|
| Animation |
Millions / High |
24.4 |
=1 |
|
| PNG |
Millions / Best Filter |
10.8 |
=1 |
|
| JPEG 2000 |
Millions / Best |
24.5 |
=1 |
|
| JPEG 2000 |
Best Depth / Best |
24.5 |
=1 |
|
| JPEG 2000 |
Millions / High |
12.3 |
=1 |
|
| JPEG 2000 |
Best Depth / High |
12.3 |
=1 |
|
| JPEG 2000 |
Millions / Medium |
11.2 |
=1 |
|
| JPEG 2000 |
Best Depth / Medium |
9.7 |
=1 |
|
| JPEG 2000 |
Millions / Low |
7.4 |
=1 |
|
| JPEG 2000 |
Best Depth / Low |
5.7 |
2 |
|
| Animation |
Thousands / Best |
13.4 |
3 |
A good, safe choice - space permitting |
| Animation |
Millions / Medium |
16.9 |
4 |
|
| Photo JPEG |
High |
5.5 |
5 |
|
| JPEG 2000 |
Millions / Least |
2.7 |
6 |
The best quality/size compromise |
| Aimation |
Millions / Low |
13.4 |
7 |
|
| Animation |
Thousands / High |
9.5 |
8 |
Banding effects visible in shadow areas |
| MPEG4 Video |
High |
6.4 |
9 |
Slightly "smeared" appearance |
| Apple Intermediate |
Other |
5.6 |
10 |
|
| H.264 |
High |
3.1 |
11 |
This popular compressor doesn't
work well with synthetic images! Definite
blurring of detail and loss of colour fidelity |
| Apple Pixlet |
Medium |
3.8 |
12 |
|
| Animation |
Millions / Least |
10.2 |
13 |
|
| Animation |
Thousands / Medium |
7.3 |
14 |
|
| H.264 |
Medium |
1.7 |
15 |
Blurred and blotched colours |
|
| MPEG4 Video |
Medium |
2.0 |
16 |
Quality unacceptable below this point |
| Photo JPEG |
Medium |
2.6 |
17 |
|
| JPEG 2000 |
Best Depth / Least |
1.7 |
18 |
|
| DVCPRO PAL |
Progressive |
8.2 |
19 |
|
| DVCPRO50 PAL |
Progressive |
16.5 |
20 |
Proof that bigger isn't always better! |
| Animation |
Thousands / Low |
5.9 |
21 |
|
| Animation |
Thousands / Least |
5.9 |
=21 |
|
| DV - PAL |
Progressive |
8.2 |
22 |
|
| Apple VC H.263 |
Best |
2.3 |
23 |
|
(*) Subjective quality ranking, based on visual inspection
of each video, where 1=best.
Discussion
We should begin our discussion with a very-important
caveat. These tests were
specific to CrystalMaker - and one structural model in
particular. They should not be interpreted in isolation.
Using different models will give different results.
In attempting to rank the different codecs, we put
particular emphasis on the appearance of video artifacts,
such as stripes or banding. These were felt to be more
objectionable than subtle blurring of fine detail - at
least, from the perspective of having a model rotating, where
spots, stripes and bands are clearly evident, whereas the
odd blur here and there is easily missed.
|
|
Examples of output quality for different
compressor settings used.
|
|
|
|
|
Animation compressor at maximum quality
|
Animation compressor at lower quality showing
severe banding in shadow areas plus horizontal streaks
across the empty black region.
|
| |
|
|
|
H.264 compressor at Medium quality.
Note the slight blotchiness and lack of
colour detail.
|
DV PAL compressor, showing general
blurriness and horizontal striping.
|
In the tests we carried out, output file sizes varied from 1.7 to
29.7 MB - that's a factor of 15 difference! The most
interesting observation, however, was that "bigger" wasn't
necessarily better. In particular, the video codecs used by
digital video (DV) cameras fared particularly poorly in
these tests - resulting in large file sizes with very-poor
video quality. Also poor was the much-vaunted H.264
compressor: clearly this is better suited to more naturalistic
scenes.
The best performing codec in these tests was the JPEG 2000
codec, and we'd recommend the Millions of Colours setting,
with the output quality set to Low or even Least.
Having said this, my hunch is that models with more straight edges - such as
polyhedral plots - would probably work better with the
Animation compressor, set to Low or Medium
quality, using Thousands of Colours. We'll have to
save this for a possible future test.
Conclusion
In conclusion, we've found that for a typical
CrystalMaker structure, the best compressor to use is
probably the JPEG 2000 compressor, set to
display Millions of Colours and using a Low
or Least quality output. Other models may require
slightly-different settings, but this table of results
should give a good general guide as to which compressors
work best
|
© 2009 CrystalMaker Software Limited. All rights reserved worldwide.
|
|
|
|