% Images with only vertical or horizontal features are preserved
% when performing 'no-op" with EWA distortion.
%
+% EWA LanczosSharp does not satisfy this condition exactly, but it
+% comes really close: If all the data is in [0,1] and the input image
+% is constant on columns (or rows), EWA LanczosSharp preserves the
+% image under no-op to within 0.00465 (that is, it is off by at most
+% 2 in 8-bit); EWA Lanczos, on the other hand, preserves it to within
+% 0.0174 (5 in 8-bit). The preservation of arbitrary images under
+% no-op, however, is not so good: EWA Lanczos preserves it to within
+% 0.308 (79 in 8-bit), EWA LanczosSharp to within 0.315 (81 in
+% 8-bit). In particular, 'pixel-hash' patterns (the checkerboard
+% mode) are smoothed out a lot.
+%
% The Lanczos2 and Lanczos2Sharp filters are 2-lobe versions of the
% Lanczos filters. The 'sharp' version uses a blur factor of
% 0.958027803631219, again chosen because the resulting EWA filter
-% comes as close as possible to satisfying the above condition.
+% comes as close as possible to satisfying the above
+% condition. (Comment from Nicolas: this value may change slightly.)
%
% Robidoux is another filter tuned for EWA. It is the Keys cubic
% filter defined by B=(228 - 108 sqrt(2))/199. Robidoux satisfies the
% "'No-Op' Vertical and Horizontal Line Preservation Condition"
-% exactly. Also, it minimally blurs high frequency 'pixel-hash'
-% patterns in the 'No-Op Distort' case. Robidoux turns out to be
-% close to both plain the Mitchell and Lanczos2Sharp filters. For
-% example, its first crossing is at (36 sqrt(2) + 123)/(72 sqrt(2) +
-% 47), almost the same as the first crossing of Mitchell and
-% Lanczos2Sharp.
+% exactly, and it moderately blurs high frequency 'pixel-hash'
+% patterns under no-op. It turns out to be close to both Mitchell
+% and Lanczos2Sharp. For example, its first crossing is at (36
+% sqrt(2) + 123)/(72 sqrt(2) + 47), almost the same as the first
+% crossing of Mitchell and Lanczos2Sharp.
%
%
% 'EXPERT' OPTIONS: