By Brad Timerson

With Avisynth (www.avisynth.org), LiMovie can read a much larger set of AVI files. Before
discovering Avisynth I used VirtualDub to convert my DV based AVI files into uncompressed AVI
and then opened the uncompressed AVI file with LiMovie. This step was necessary because I do not
have a DV codec that works with the Video for Windows API and LiMovie can only read AVI files
via the Video for Windows interface. With Avisynth installed I can now skip this step and work
directly with the DV based AVI files. With this new approach I can now examine much longer time
spans with LiMovie (useful for lunar occulations) and it saves the time (and space) consuming step of
generating an uncompressed version of the AVI file. This web page provides a brief tutorial on using
Avisynth with LiMovie and includes some notes on some of the esoteric features.
To use Avisynth: install it, create a text "script file" with the extension .avs, and open this script file
with LiMovie (or VirtualDub).. The script file is a simple text file that includes commands telling
Avisynth which AVI file to read, how to read it, and optionally provides for some filtering or
enhancements. To utilize Avisynth with LiMovie you open the *.avs script file with LiMovie (instead
of opening the AVI file itself). When LiMovie opens the *.avs file, Avisynth reads the specified AVI
file, interprets the file data with the appropriate codec, and sends the data to LiMovie. Why do
this? Because Avisynth can utilize the DirectShow video subsystem and thereby read just about any
AVI compression. Unlike LiMovie, Avisynth is not limited to the Video For Windows codecs
installed on your machine - it can utilized any codec. In addition Avisynth includes several useful
image enhancement filters, and there are many additional plug-ins provided by Avisynth users for
reading difficult or unusual video formats.
For example, let's say you would like to review the DV based video file "c:\videodata\occultation.avi"
with LiMovie. Create a text file (with Notepad) called occultation.avs (could be any name with an .avs
extension) and enter the following line in this file.
DirectShowSource("c:\videodata\occultation.avi")
Then start LiMovie and open occultation.avs. Upon opening occultation.avs, Avisynth will interpret
the DV data (using the MS default DV codec for DirectShow) and feed it to LiMovie.
Note: In my current version of LiMovie the file open dialog will only browse for *.avi files (not *.avs
or *.* files), but I can type in the filename directly - occultation.avs and it works.
Actually Avisynth has more options than DirectShow for reading AVI files (e.g. AVISource,
OpenDMLSource, ...). I used DirectShow because my DV files are type-1 and only DirectShow would
read those. You may want to review the Avisynth documentation to determine which function is most
appropriate for your situation. There are many sources of information on Avisynth on the net. As of
this writing the avisynth website has been unavailable for several days, so I have posted a recent
version of Avisynth (version 2.57) on my site: Avisynth self-extracting archive (*.exe file). To install
Avisynth, download this file and run it. It will install Avisynth and create a program group in your
start menu (including a link to some documentation).
Levels Filter - Basic Image enhancement
Avisynth includes a built in filter called "levels" that is analogous to the levels command in many
image manipulation programs (including VirtualDub and Photoshop). I find that this filter is often
useful when viewing videos containing dim stars. In these videos I use the level command to stretch
out the contrast for the darker parts of the image by lowering the top end of the input range and often
adjusting the "gamma" or mid-tone adjustment as well. AviSynth is not well suited to previewing
these adjustments because you must iteratively change the script file then re-open script file to view the
changes. Instead I open the video file with VirtualDub, add the levels filter and use the preview mode
of this filter to determine the best settings. Then I input these settings into the Avisynth levels
command in the *.avs script file. Here is an example of a script that makes a dim star more visible.
DirectShowSource("c:\videodata\occultation.avi")
Levels(0,0.7,50,0,255,coring=false)
Note: the coring argument is related to the issue of the "color space" of the data in the AVI file. More
specifically, this relates to the range of luminosity and chroma values for the pixels in a frame. Video
destined for TVs should meet the CCIR-601 criteria for digital video and this limits the numerical
values for black to a luma value of 16 and white to a luma value of 235. However, there are situations
where a file may contain values outside these ranges. For example, I have found that the DV files that
I download from my Sony camera contain pixels with values outside the "standard" range. I don't
know if it is the camera or the capture software but it is there. If you do not specify "coring=false" in
the levels command, the Avisynth Levels function will rudely clip all values less than 16 to 16 and all
values greater than 235 to 235. In most situations this is not a problem because even dim stars seem to
be above 16 and it is rare that we would want to recover data from the "super-white" range of 235 to
255. However, I recommend using coring= false as a default. If you wish to later adjust your image to
get rid of any 'super-black' and/or 'super-white' values, use the command "ColorYUV(levels="pc-
>tv"). This command will do a more intelligent job of mapping the colors from the 0,255 range into
the 16,235 range (better than a simple clipping approach). For more information see the comments in
the Avisynth documentation and this description of "The White-clip problem".
Running Average Script
As an exercise and a demonstration of the flexibility of Avisynth I created a script that adds together
frames in a video as a running average. The purpose of this filter is to increase the signal to noise of
very dim stars to make them more "visible" to LiMovie (and the eye). Of course, this approach
reduces the time resolution of the video but it may be the only option with the star is dim. This script
is not "smart" enough to track the motion of a star. If the star is moving around in the span of the
frames averaged the star will be "blurred" across the image. This script allows for a "box" of the image
to contain a timestamp (e.g. Kiwi OSD time insertion).
Although the time resolution is now down to 3 * 1/30 second = 1/10 second you can actually see some
stars. By adding a Levels filter you can make these stars even more apparent. This is quite an
improvement.
Here is the script that I used. Note that I have defined two script functions: AddThree and AddFive at
the beginning of the script. At the end of the script the AVI is opened and the AddThree function is
called.
<begin script file AddFrames.avs >
#
# AddThree - produces a three frame running average version of the input AVI (to reduce noise)
#
# This function adds (not averaged) together the frames of a clip as a running average (three frames at
a time).
# The frames are added, not averaged so this function will "blow out" any bright values. The middle
frame of the three
# is assumed to be the "time frame" and the time label for this frame is overlayed on the final
"summed" frame. The
# location of the time label in the frame is specified via the X1,Y1,X2,Y2 coordinates which are
analagous to the
# left, top, width/-right, height/-bottom coordinates specified for the Avisynth built-in function "Crop".
#
#
# Inputs:
# c = the input clip
# X1,Y1 = x,y coordinates of the upper left of the time label
# X2,Y2 = either width,height or -right,-bottom coordinates for the time label
#
#
#
#
Function AddThree ( clip c , int X1, int Y1, int X2, int Y2) {
intFrames = Framecount(c)
clipL1=Trim(c, 0,intFrames-3)
ClipM=Trim(c, 1,intFrames-2)
ClipR1=Trim(c, 2,intFrames-1)
ClipLabel=Crop(c, X1,Y1,X2,Y2)
Overlay(ClipL1,ClipR1, x=0, y=0, mode="add", opacity=1.0)
Overlay(ClipM,x=0,y=0,mode="add",opacity=1.0)
Overlay(ClipLabel,x=X1,y=Y1,mode="blend",opacity=1.0)
return (Trim(0,intFrames-1))
}
#
# AddFive - produces a Five frame running average version of the input AVI (to reduce noise)
#
# This function adds together the frames of a clip as a running average (five frames at a time).
# The frames are added, not averaged so this function will "blow out" any bright values. The middle
frame of the five
# is assumed to be the "time frame" and the time label for this frame is overlayed on the final
"summed" frame. The
# location of the time label in the frame is specified via the X1,Y1,X2,Y2 coordinates which are
analagous to the
# left, top, width/-right, height/-bottom coordinates specified for the Avisynth built-in function "Crop".
#
#
# Inputs:
# c = the input clip
# X1,Y1 = x,y coordinates of the upper left of the time label
# X2,Y2 = either width,height or -right,-bottom coordinates for the time label
#
# Example:
# DirectShowSource("C:\Documents and Settings\Steve\My Documents\Ulead
VideoStudio\Occultations\Archive\dimtest.avi")
# AddThree(Last,80,50,-96,-285)
#
#
#
Function AddFive ( clip c , int X1, int Y1, int X2, int Y2) {
intFrames = Framecount(c)
ClipL2=Trim(c, 0,intFrames-5)
ClipL1=Trim(c, 1,intFrames-4)
ClipM=Trim(c, 2,intFrames-3)
ClipR1=Trim(c, 3,intFrames-2)
ClipR2=Trim(c, 4, intFrames-1)
ClipLabel=Crop(c, X1,Y1,X2,Y2)
Overlay(ClipL2,ClipL1, x=0, y=0, mode="add", opacity=1.0)
Overlay(ClipM, x=0, y=0, mode="add", opacity=1.0)
Overlay(ClipR1,x=0, y=0, mode="add", opacity=1.0)
Overlay(ClipR2,x=0, y=0, mode="add", opacity=1.0)
Overlay(ClipLabel,x=X1,y=Y1,mode="blend",opacity=1.0)
return (Trim(0,intFrames-4))
}
#
# Now open the AVI file and call the averaging function
#
DirectShowSource("C:\videotest\dimtest.avi")
AddThree(Last,80,50,-96,-285)
<end of script file>