I’m using ImageMagick to downsample text. I realize that there’s no such thing as the best filter for every situation, but I’m thinking there must be a generally-accepted standard when it comes to downsampling text specifically (I could be wrong). Here’s the list of filters available in ImageMagick:
- Bartlett
- Blackman
- Bohman
- Box
- Catrom
- Cubic
- Gaussian
- Hamming
- Hanning
- Hermite
- Jinc
- Kaiser
- Lagrange
- Lanczos
- LanczosSharp
- Lanczos2
- Lanczos2Sharp
- Mitchell
- Parzen
- Point
- Quadratic
- Robidoux
- Sinc
- SincFast
- Triangle
- Welsh
There is no generally-accepted downsampling filter; simply because, there’s no generally-accepted type-face for text. You’ll need to identify the font(s) your working with, and apply the best-matching filter for said task.
Visual comparision of each filter’s stangth & weakness can be found here & here.
I like to remind myself which filter will meet a task by generating a visual cheat-sheet.
Here’s a quick bash script previewing “DejaVu Sans Condensed”