Steven Silvester
b710bd21d2
Use a stacklevel of 2 by default for all warnings
2016-01-30 08:57:46 -06:00
Gregory R. Lee
682f0895e7
TST: add test for constent TV weighting across number of image dimensions for Chambolle algorithm
2016-01-24 12:27:10 -05:00
Gregory R. Lee
53742be9c6
fix axis bug in multichannel case and add a corresponding test
2016-01-24 02:24:37 -05:00
Gregory R. Lee
a30592a22a
MAINT: explicitly specify axis argument
2016-01-23 11:23:21 -05:00
Gregory R. Lee
e6a1a19337
update default weight to reflect the bugfix from hardcoded 0.5 to 1/(2*ndim)
...
use the default weight in all the tests as well.
2016-01-23 11:01:48 -05:00
Gregory R. Lee
dd5a708a33
minor bugfix so that weight gives a consistent amount of smoothing regardless of the number of dimensions.
...
As in the original reference, the stepsize should be in both the numerator and denominator of p.
2016-01-23 11:01:15 -05:00
Gregory R. Lee
9f419024ad
Switch _denoise_tv_chambolle to use C-ordering
2016-01-23 10:59:33 -05:00
Gregory R. Lee and Gregory R. Lee
4a8a2189f2
use Fortran order for g and p in _denoise_chambolle_nd to avoid a performance regression
2016-01-23 10:59:32 -05:00
Gregory R. Lee and Gregory R. Lee
4bc4f0a169
remove try/except in _denoise_chambolle_nd. just use the numpy 1.6 compatible case
2016-01-23 10:59:32 -05:00
Gregory R. Lee and Gregory R. Lee
4a0ad77780
update tests for denoise_tv_chambolle
2016-01-23 10:59:32 -05:00
Gregory R. Lee and Gregory R. Lee
c2b86d8a02
fix numpy 1.6 compatibility
2016-01-23 10:59:32 -05:00
Gregory R. Lee and Gregory R. Lee
2d326a464e
ENH: n-dimensional refactor of TV denoising.
...
_denoise_tv_chambolle_2d and _denoise_tv_chambolle_3d are replaced by _denoise_tv_chambolle_nd
The restriction to 2D and 3D in denoies_tv_chambolle was removed.
2016-01-23 10:59:32 -05:00
Juan Nunez-Iglesias
e53d7daaa3
Merge pull request #1804 from soupault/inpainting
...
ENH: Inpainting with biharmonic equation
2016-01-18 10:56:24 +11:00
Egor Panfilov
3ae2f1236b
TST, DOC: Added test for 3D spatial data, cleaned up gallery example
2016-01-16 20:02:24 +03:00
Egor Panfilov
98fec5211a
FIX: Returned handle for empty input arrays
2016-01-15 20:04:08 +03:00
Egor Panfilov
893d937491
ENH: Moved nested functions, refactor code
2016-01-15 13:59:32 +03:00
Emmanuelle Gouillart
33a2182e06
Merge pull request #1705 from deledalle/ref_nlmeans
...
DOC : add ref to nlmeans
2015-12-30 22:06:19 +01:00
Egor Panfilov
89784631e7
ENH: Massively reworked code to enable nD-support
2015-12-30 22:45:27 +03:00
John Wiggins
72f542100e
Use np.empty() instead of malloc/free in _denoise_bilateral.
2015-12-27 23:29:59 -06:00
John Wiggins
e528d47d5b
FIX: Move a argument check to avoid leaking malloc'd arrays.
2015-12-26 10:55:28 -06:00
Egor Panfilov
0cc4402c19
DOC: Fixed docstring, moved to assert_nD
2015-12-24 20:35:32 +03:00
Egor Panfilov
2fcf2eafe7
ENH: Added support for multichannel arrays, fixed float scale to [0;1]
2015-12-24 20:35:32 +03:00
Egor Panfilov
0e6e08d0c0
TST: Attempt to fix Travis bug with old packages
2015-12-24 20:35:32 +03:00
Egor
92f3114c9e
FIX: Fixed test for biharmonic in accordance to new kernel for edges
2015-12-24 20:35:32 +03:00
Egor
add5b8b4d6
FIX/TST: Fixed bug with defected bounding frame, added testcase
2015-12-24 20:35:32 +03:00
Egor Panfilov
1f3721fcbe
ENH: Inpainting with biharmonic equation
2015-12-24 20:35:32 +03:00
Emmanuelle Gouillart
6740ef9849
Merge pull request #1842 from jni/1840
...
Specify 2D images in denoise bilateral docstring
2015-12-24 09:59:11 +01:00
Gregory R. Lee
3f5aa73eb4
Update tests to also use reasonable weights for TV denoising
2015-12-23 12:30:12 -05:00
Gregory R. Lee
f143f4b39e
Change default weight for denoise_tv_chambolle to a value that is reasonable for float images scaled to the [0, 1] range
2015-12-23 12:29:37 -05:00
Juan Nunez-Iglesias
e102011098
Specify 2D images in denoise bilateral docstring
...
Addresses #1840
2015-12-23 23:57:22 +11:00
scottsievert
fb99e3cce6
styles code (no more functions, less lines)
2015-12-15 19:19:14 -06:00
scottsievert
2892e90abc
removes paren
2015-12-13 16:59:59 -06:00
scottsievert
4d6c9e7b84
now N dimensional, changes constant, cleans and comments
2015-12-13 14:57:52 -06:00
Charles Deledalle
cc24efbe2b
DOC : change sneaky unicode in the endash
2015-12-13 18:58:01 +01:00
scottsievert
3fd47091dd
adds comment explaining time_ratio decision
2015-12-12 21:15:01 -06:00
scottsievert
3f3b705ebd
fixes import, tuple addition
2015-12-12 19:26:10 -06:00
scottsievert
1be8336599
adds decision when to use fftconvolve/convolve2d
2015-12-10 19:47:02 -06:00
scottsievert
5aabc199aa
uses fftconvolve instead of convolve2d
2015-11-30 23:25:44 -06:00
Charles
e822e5a556
DOC : add ref to nlmeans in the API
2015-09-22 11:14:33 +02:00
Charles
8c53663dc1
DOC : add ref to nlmeans
2015-09-11 16:31:03 +02:00
Johannes Schönberger
a664043197
Merge pull request #1667 from odebeir/sprint_euroscipy2015_denoise
...
Sprint euroscipy2015 denoise
2015-09-04 08:49:09 -04:00
Olivier Debeir
f5e4a5e44d
remove import/blank lines
2015-09-02 16:01:13 +02:00
Emmanuelle Gouillart
fa9e000ae4
Merge pull request #1631 from stefanv/rename_denoise_nl_means
...
Rename denoise nl means
2015-08-29 17:33:05 +02:00
odebeir
544fb95f76
typo
2015-08-28 21:35:18 +02:00
odebeir
76285f5986
add exemple to denoise_bilateral() in docstring
2015-08-28 21:33:58 +02:00
Arnaud De Bruecker
50804d1e9b
Correct documentation to denoise_bilateral for sigma_range
...
Closes #1630
2015-08-28 19:43:14 +01:00
Arnaud De Bruecker
955b1761f1
Add documantation to denoise_bilateral for sigma_range
...
closes #1630
2015-08-28 19:10:09 +01:00
Stefan van der Walt
e084c5fd50
Rename nl_means_denoising to denoise_nl_means
2015-08-27 13:19:02 -07:00
Johannes Schönberger
236aa208fe
Merge pull request #1583 from grlee77/fix_interp_modes
...
FIX: bug in 'reflect' and 'wrap' coordinate mapping
2015-08-20 18:53:11 -04:00
Gregory R. Lee
81ea7a6e34
MAINT: All modes in _shared.interpolation.pxd were changed to be consistent with numpy.pad naming conventions. Specifically 'nearest' was changed to 'edge' and 'mirror' was changed to 'reflect'. All functions with a mode argument that rely on these functions had their inputs changed accordingly. For now there is a deprecation warning if the user supplies mode 'nearest'. Mode 'mirror' never appeared in an official release of skimage and so has no corresponding deprecation warning.
2015-08-13 23:39:55 -04:00
Steven Silvester