c0ed106079
Formatting and whitespace changes
2020-06-06 14:25:03 +02:00
d64385ae99
revise kernel mask calculations
...
We were seeing small displacements at some shrink factors because of
rounding in the mask size calcualation. This PR takes the rounding into
account when positioning the mask.
See https://github.com/libvips/libvips/pull/1592#issuecomment-626363031
2020-05-10 18:45:27 +01:00
29d9673fd8
update comment
2019-05-14 08:20:30 +01:00
3b1e3e1841
add Mitchell kernel
...
see https://github.com/jcupitt/libvips/issues/966
2018-06-12 12:55:10 +01:00
a8bbe05596
tiny optimisations
2017-10-16 20:55:58 +01:00
0767e4f4c9
remove a 0.5 pixel shift from triangle
...
lines up with cubic and lanczos now
2017-10-16 09:38:04 +01:00
68ed42e2fa
make linear and cubic adaptive
2017-10-15 22:09:44 +01:00
65b9ced6e9
improve arg order
...
more consistent
2016-04-11 13:23:25 +01:00
269cbb8641
better mask sizing for lanczos
2016-04-11 13:19:00 +01:00
c2035eb0f0
similarity uses reduce more often
2016-03-21 14:18:01 +00:00
05156edd15
add orc path for reducev
...
works, but still not quite right, see some banding
2016-03-18 17:41:51 +00:00
ff88087a28
try making the reduce mask larger with residual
2016-03-15 14:59:59 +00:00
4d10bd12f9
still trying to get reducevl3 to vectorise
2016-03-10 10:24:44 +00:00
a26291a426
reducehl3 is done
...
sort of anyway, tinkering with optimization
2016-03-09 22:47:02 +00:00
e9728e389b
first commit
2016-03-09 12:56:30 +00:00
8cffd136e9
finish reduce
...
and fix up bicubic a bit, it works better on int32 images now
2016-02-07 17:41:39 +00:00
3e78bdb6a9
reduceh is working!
2016-01-27 14:34:06 +00:00
8b09c86492
add cubich interp., start reduceh
2016-01-27 11:12:36 +00:00
25eeb82de6
Consolidate FAST_PSEUDO_FLOOR macros, prefer __builtin_floor if available
...
Use of vips_interpolate_get_method avoids need to pass interpolate instance
Remove an unnecessary temp variable to help prevent register starvation
5% less time in vips_interpolate_bicubic_interpolate
10% less time in vips_affine_gen
2016-01-22 22:26:29 +00:00
2b086cfe6e
add signed int bicubic path
...
better rounding for 8- and 16-bit signed int bicubic
2014-08-07 10:13:08 +01:00
ebeee822bc
better rounding behaviour for bicubic
...
fixed-point bicubic was not rounding to nearest, introducing some noise
in flat areas
2014-08-05 10:02:11 +01:00
8cc76a6ba3
Update address for Free Software Foundation
...
Also update LGPLv2.1 license text from the copy currently on the GNU
website.
2013-03-07 00:40:19 -05:00
f27ab2fa3f
fix window_offset stuff, fix a performance problem
2010-08-12 15:56:07 +00:00
3545786e81
tweak of the 5*, 1+ and 5- computation of the four Catmull-Rom weights
2010-06-10 19:54:27 +00:00
736836a099
computation of Catmull-Rom weights with 10 flops: has got to be a world record
2010-06-10 19:49:50 +00:00
06cee27b65
removed outdated minmod from templates (only used by nohalo.cpp now anyway)
2010-05-30 20:48:08 +00:00
d0573100b3
interpolation precision hacking
2010-05-28 21:16:28 +00:00
b0b7dc43e3
added mbicubic = symmetrized monotone cubic splines and added hopefully fast cast operators in templates.h
2010-02-13 01:46:00 +00:00
4e76695e19
merge gtkdoc branch
2009-08-16 15:00:08 +00:00
Kleis Auke Wolthuizen
John Cupitt
Lovell Fuller
Benjamin Gilbert
Nicolas Robidoux