Video Compression Options... Revisited
As movie editors, we live in a world
filled with different codecs... and codec-related issues. This newsletter
is another step toward becoming more versed in and comfortable with
them.
It'll take you through a couple
tests, first checking the rendering times and file sizes by using all the
available compressor options on my new HP notebook, and then
checking differences in visual quality.
The tests will be done
using default compressor settings.
We'll also take a look at the
customizable settings for the Divx and Windows Media 9 compressors and
do a special check of enhanced videos using two-pass renderings for each.
... before getting into it, here
are a couple notes...
Notes...
Custom Overlay
Projects is the subject of a new section on the Editing > Text
> Custom Overlays page of the website
.
In addition to
the
puzzle project made for newsletter #73, I made a couple
more.
A
Keyboard project
started on Christmas. One of our grands got a keyboard as a present, and
I thought it would be neat to record some music from it
with my camcorder. It was playing some
samples by itself but the keys weren't moving, so
I took a snapshot from the video and used it for something more
interesting.
... another overlay project is a
Countdown
Clock-like one, with 10 assorted watch faces being removed as
the seconds tick down.
The website page includes
instructions about setting up and using the projects, and links
to downloadable packages with everything you need.
Each package has
a project MSWMM file, an xml file, a set of transparent png images,
and batch files to install and remove them. It's a good way to explore the
use of custom xml files without having to write or tweak them. Those who enjoy
getting into them deeply can go as far as they want.
.... on to
the main topic...
some
Video Codec testing...
It's been a
while since I tested all the codec options on my computer... and made
a table of the results such as the one on the Importing Source Files
> Video > Recorded TV page of the site. I've pointed lots of users to that
table since it first rolled it out.
That table was done with
MPEG-2 files as inputs. I'll do this issue using AVI files, one a random
AVI from a magazine CD, and the other a DV-AVI made by Movie Maker 2.
Codecs change with time...
some go, new ones emerge, some are revised... and your
computer changes. For this round of tests, I'll use my new HP
Pavilion zd8000 notebook with its 3.4 GHz CPU...
The computer currently
has 20 compression choices in the drop down lists of video editing apps and
utilities. The full list of them is in the table below.
For a user perspective, it's
not about all of the codecs on my system per GSpot or other utilities...
it's about which compressors show up in the picklist of choices when I use
an app, and how well they work when I select them.
I'll use VDubMod to do
the renderings.
The first source
file is a 17+ second AVI file from a CD in a magazine... with
a file name of USSConstellation.avi. Its properties are: 520x376
pixels, 24.988 fps, 17+ seconds, compressed with the Microsoft Video 1
codec, no audio stream. 27.9 MB file size.
The frame snapshot at
the left shows that the video includes black bars at the top/bottom and the
right side... the quality isn't high and the video ends with the last frame
offering the longer higher quality clip for sale. It'll do as an input file to
check rendering times and file sizes.
Rendering Times and
File Sizes...
Let's keep the same frame size
of 520x376 pixels, not crop away the black borders, and use whatever
default settings the compressors have... to see which ones work, how long the
renders take, and how large the new file is.
Here's the tally.
The rendering times that were faster than the 17+ second real time
duration are on the left side of the time column. Those whose file sizes
were smaller than the original are on the left size of the file size
column.
The ranges of times and sizes
are considerable. Some were noticeably fast, and some resulted in smaller
file sizes. The 4 compressors in bold type were the only ones that
were both quick and produced smaller files.
|
Compressor |
time (seconds) |
file size (MB) |
| |
|
|
|
|
18 |
258.7 |
|
Cinepak Codec by Radius |
52 |
10.6 |
|
DivX Codec
4.12 |
5 |
1.8 |
|
Indeo video 5.10 |
40 |
10.8 |
|
Intel 4.2.0 Video V2.50 |
N/A |
N/A |
|
Intel Indeo(R) Video
R3.2 |
26 |
7.7 |
|
Intel Indeo Video 4.5 |
86 |
13.4 |
|
Intel IYUV codec |
8 |
129.4 |
|
Microsoft H.261 Video
Codec |
N/A |
N/A |
|
Microsoft H.263 Video
Codec |
N/A |
N/A |
Microsoft MPEG-4 Video
Codec V1 |
6 |
6.7 |
|
Microsoft MPEG-4 Video
Codec V2 |
6 |
6.5 |
|
Microsoft RLE |
N/A |
N/A |
|
Microsoft Video 1 |
7 |
19.2 |
|
Microsoft Windows Media Video
9 |
26 |
1.5 |
|
Panasonic DV CODEC (resizing to
720x480) |
7 |
52.9 |
PCLEPIM1 32-bit
Compressor |
N/A |
N/A |
|
PCLEPIXL 32-bit
Compressor |
8 |
86.2 |
|
PICVideo MJPEG
Codec |
4 |
7.1 |
|
Zoran Decompressor
|
N/A |
N/A |
Using the Panasonic DV
without resizing to standard DV-AVI size of 720x480 gave an error:
"Cannot start video compression: The source image format is not acceptable
(error code -2)". I added the resize filter to get past this test.
The 6 compressors with
N/A in the time and size columns didn't. The message was: "error getting
compressor output format size".
I imported the 14 new files
into Movie Maker, previewed them in the collection, put them on the
timeline, previewed the project, and rendered a movie... there were no
issues.
Visual
Quality...
The test renderings above
started with a lower quality video, which set the stage for all of the files to
look similarly poor. To equalize the playing field for the visual
comparisons, I used a different higher quality input file.
-
I made a Photo Story from
a 6+ megapixel still picture, saving it at 800x600 pixels. Movie
Maker converted the story into a DV-AVI file. VDubMod made
a set of AVI files with the different compressors, and saved frame #390
from each as a BMP image.
This copy of the collage shows
the frame slices at actual sizes, a JPG file at 80% quality... if
you're interested in seeing the 200% sized full quality 9.7 MB BMP
file,
here's
a link. I recommend downloading it and taking a look.
The only thing added to the
frame slices for this collage is the name of the compressors in black at
the lower left; the other script was added to the original image before
importing into Photo Story.
If at first glance you think
they all look pretty good, I agree with you! Looking closer, it's easier to
select losers such as the Motion JPEG compressor slice at
the bottom of the collage... but it's too hard
to pick a winner... and probably not fair to judge the overall video
by looking at a small slice of one frame. After all, it's a movie, not a still
picture.
Two of the codecs which
made it through the initial rendering test didn't make it to through
this step for visual comparisons. I got error messages when trying to
use these two with the DV-AVI input file. They don't show up on my older
list online either.
Intel Indeo(R) Video R3.2
Intel IYUV
codec
Enough of them made it through both tests
that I didn't stop to study why these two didn't.
Codec Settings...
The two rounds
of tests above were done with default compressor settings. Many
of them have adjustable settings.
As the settings
are used to balance quality, rendering speed, and file sizes, tweaking
them and doing more testing could make this a full-year project
instead of a few hours for a weekly newsletter. Let's take a general look
at two of them.
Divx
4.12 Settings...
Here's the
configuration window for the Divx 4.12 codec. You get to it when selecting the
codec from the drop down list and pressing the 'Configure' button. For info
about the settings, use the help file available from that
button.
I did a 2-pass
rendering to a high 6000 kbps bitrate to see how the saved video differed from
the default in the collage.
The rendering
time was much longer and the file size 4-1/2 times that of the rendered file
with the default settings. Here's the same slice
of the 390th frame, with the enhanced quality clearly showing
that it's a good reason to go into
compressor settings.
Windows
Media 9 Settings...
Not to be
outdone by the tweaked Divx compressor, here's the slice from a file
rendered with equally tweaked compressor settings, a 2-pass file using
the higher 6000 kbps bitrate.
The Compression
tab of the Windows Media Video 9 settings window shows the
default is a one pass quality based VBR file with the quality
level set to 80.
Pressing the Help button gets you lots to study... for example this
paragraph about the Quality Level setting:
Quality level
The quality level specifies the quality of the
compressed content that the encoder should maintain when using quality-based
VBR. This value ranges from 0 to 100, with 100 being the highest
quality.
Not all of the values in the range have a unique
meaning. The values that represent a step up in quality from the previous level
are: 1, 4, 8, 11, 15, 18, 22, 25, 29, 33, 36, 40, 43, 47, 50, 54, 58, 61, 65,
68, 72, 75, 79, 83, 86, 90, 93, 97, and 100.
A few
comments about these two test rounds:
The fastest
render was with the PICVideo MJPEG Codec, taking only
4 seconds versus the longest time of 86 seconds, but
it's a loser when it comes to visual quality (it's the
bottom slice of the collage).
The smallest file
size of 1.5 MB goes to the Windows Media 9 codec, just
over 5% the size of the original file. But without tweaking the settings, it
doesn't win the visual quality check either.
The Uncompressed RGB file is
almost 10 times the size of the original. It maintains the quality but can't
enhance it.
Of the 4 compressors
tagged in the table above for giving both fast results and small
file sizes, the two Microsoft MPEG-4 compressors held in there
for the quality rankings... the other two (Divx and MJPEG) ranked lower on
the visual collage.
Conclusions and Closing
Of the hundreds of codecs and
filters on my system, some are used to decompress files and not compress them,
some are locked from anything but use with the software they came with, some
have restrictions for input files.... etc, etc. The world of codecs have
a lot more complexity than most users of Movie Maker, including
myself, want to get into. We usually just want to know which one to use and
how to use it.
What's
different since the similar testing for the importing
recorded TV page of the website? For starters, I'm doing the tests on a
different computer. Most of the codecs are the same. 9 compressors
made it to the list the first time I did this testing, and 11 this time... PCLEPIXL and PICVideo MJPEG are new
to the list... they came with Pinnacle Studio
8.
I don't know if I'm being fair to the
Divx compressor, as I've used the version 4.12
for years... newer ones might be better
but I'm cautious about the spyware, adware, or worse that can
come with Codec packages. The version continues to work for me so
I'll continue with it... and it did great in the extra test, the first time
that I tweaked Divx compressor settings.
A true visual
quality test would have all compressor settings tweaked for
optimal performance. As utilities such as VDubMod play a secondary
role to Movie Maker, there's no reason to put the effort into doing such
testing. The important thing to remember is that if, for whatever reason, you
find yourself using one of these compressors for a conversion... get into the
settings and make the new file the best quality you can. The two extra tests I
did with the Divx and Windows Media 9 compressors give me a better
appreciation for the extra quality you can get with minor changes
in settings. I usually prefer extra visual quality at the cost
of longer rendering times and larger files.
The Panasonic DV
compressor is still my current
favorite all-purpose codec. Its large file size is expected when
working with full quality Digital Video. Its rendering time is short, and
the output is high quality. As with the Microsoft DV codec used by Movie
Maker, there are no compressor settings to tweak... things are simpler with
DV codecs.
Have a great New Year's
Eve and New Year!!!
we'll be seeing Movie Maker for Vista in
2006
Have a great week...
PapaJohn
Movie Maker 2 and Photo Story
