I know the project doesn't get much action now, but I thought I'd ask...
Would you consider implementing something akin to Cleartype for sharper resizing? It works best for greyscale, but works pretty well even for color images. Text legibility benefits the most, image clarity less so.
The basic method is simple, use the given resize algorithm to size the image to 3x output width by 1x output height, then take every three pixels of that as [r,g,b] of the output. One option for doing this is to point resize it down three times, cropping 0, 1, and 2 pixels off, then merging the components of each, as I do in my batch file. You can do all of this in one operation by modifying initial resize parameters for each channel.
Unfortunately I have trouble obtaining Comical source over svn (a certain error is always thrown) so I'm not sure if I can help, unless it's from 0.8 codebase.