(Courtesy from: http://www.dscaler.org/Help/FAQ.htm)

What are all those deinterlace algorithms ?

Video Deinterlace (Bob)               (This is the BOB filter used in xdtv)
-----------------------
This method is based on Gunnar Thalin's VirtualDub filter.  If it detects
weaving artifacts in the current image it uses bob to get rid of them. 
This method has a tendency to bob rather too much and gives poor results on
fine static images.


Video Deinterlace (Weave)           (This is the WEAVE filter used in xdtv)
-------------------------
Similar to Video Deinterlace (Bob), however this method has a tendency to weave
on moving images.


Video Deinterlace (2-Frame)
---------------------------
This method uses the current frame and the last two to determine whether to bob
or weave a given pixel. This gives better results on both stationary and moving
images than the above two methods however it uses more CPU.


Simple Weave
------------
This method just combines the most recent even and odd fields. This causes
motion artifacts and is provided for comparison.


Simple Bob
----------
This uses only the most recent field and fills the lines in between with
interpolated pixels. This method has in may ways been superceded by Scaler BOB.


Blended Clip
------------
The BLENDED CLIP Deinterlace method is designed mostly for experimentation by
those interested in what works for video deinterlacing. When it is selected it
will pop up a control panel with many controls to adjust various parameters.
These all correspond to Blc... parameters in the [Deinterlace] section of the
DScaler.ini file and are documented there. It is probably not the single best
method for anything but it can be used to create custom deinterlace methods for
special purposes. These can be saved in a separate ini file and invoked by a
command line parameter when starting DScaler.exe. Somewhat CPU intensive


Scaler Bob
----------
Uses the DirectDraw hardware bob feature (if available), this uses the current
field and lets the video card scaler size the picture to fit the screen.
Produces good results for fast moving images.


Even Fields Only
Odd Fields Only                 (This is the ONE FIELD filter used in xdtv)
----------------
These use only one field out of each pair and use hardware scaling. This
reduces bob artifacts at the cost of loss of information. These were originally
intended to be used with game consoles.


Adaptive Deinterlace
--------------------
This method uses the amount of motion detected to select the best deinterlacing alogoritm. Currently by default it switches between 2-Frame and weave.


Greedy deinterlace method
-------------------------
The GREEDY Deinterlace method is designed to give good results on low motion
video source, using less CPU than some others. Good on slower machines but it
will give awful results on high motion video material like sports. Quite by
accident, it can also do a decent job on poorly mastered 3:2 or 2:2 pull down
film source material like some anime movies, regardless of low or high motion.
Try it if you have a movie source that is not giving good results with normal
Auto Pull Down processing.


Greedy 2 Frame method                 (This the GREEDY filter used in xdtv)
---------------------
This method tries to combine the best features of the 2-frame algorithm with
the looking ahead ability of the greedy method. I find this methods works well 
on sports and other high motion material.  With static images and low quality
sources there may be unacceptable levels of bobbing.