-0.1:
This is the first release.

-0.2:
Some code cleanup, exit reports.
Replaced resize.c with the original (were vertical stripes in black).
Need to look into this.
Added -l and -f command line flag for lowpass filter operation only.
You can specify the cutoff frequency (for PAL).
Added #define NO_GNU_C option in resizer.h, to activate readline code
using fgetc (with thanks to "Steven M. Schultz" <sms@2BSD.COM>.

calculate slope
Trying to make a much as possible a smooth cutoff, preventing 'ringing',
this is especially visible at low cutoff frequencies in black areas.

Specification of filter slope frequency:
form A to B is one half period of a sinewave of frequency dfslope,
this is what you enter with the -s  command line flag.
The -f command line flag is the frequency at which the amplitude is 1/2,
in point C.
So the filter starts a bit before that point and ends a bit after that point.
Try flags -l -f 3.5 -s 1 , this creates a low pass 50 % down at 3.5 MHz,
and rolls of smoothly to be 100 % down at 4.0 MHz.
Thats is if I did the math right.
View graph with fixed spacing!

      A                  B
. . . . 
           .
             .
              .
               .
              C .
                 .
                  . 
                   .
                      .
                           . . . .

Added -b flag, burst mask, this allows you to blank out possible color
burst ruminants (that could appear as a vertical line at the most left).
Try -b 10 for making the most left 10 pixels of each line black.
This is applied BEFORE fft and resizing.


0.3:
log option added -g
This creates a slope based on a logarithmic frequency scale.
As everything else here, this is all experimental.


0.4:
Dropped most command line options, added -v to print out the filter,
all coefficients calculated automatically, filter now only uses 
exp(-u^2 / b).
This programs ONLY uses fft low pass in horizontal(rescaling).
The program only works without to many artefacts if resizing from 720x576
DVB-s (digital satellite PAL Europe) to 352x576 CVD (China Video Disk).
ANY  other format seems to give a lot of distortion in the mjpegtools
color bar test signal (square waves).
If anyone knows a better algo, pls let me know.
Filter order is also automatically determined.

