Mailman 3 March 2017 - scikit-image

Re: [scikit-image] Announcement: scikit-image 0.13.0
by François Boulogne March 29, 2017

March 29, 2017

Thank you Juan for making it possible and congrats to the contributors, especially the new ones who made their first contributions. Best, -- François Boulogne. http://www.sciunto.org GPG: 32D5F22F

1 0

Re: [scikit-image] Announcement: scikit-image 0.13.0
by Stefan van der Walt March 29, 2017

March 29, 2017

On Wed, Mar 29, 2017, at 11:22, Juan Nunez-Iglesias wrote: > We're happy to (finally) announce the release of scikit-image v0.13.0! > Special thanks to our many contributors for making it possible! This > release is the result of over a year of work, with over 200 pull > requests by 82 contributors. I want to send a huge shout-out to Juan for the great job he did on this release. v0.13.0 was big, and he took great care in sorting out various technical issues along the way. Thanks, Juan! Enjoy, everyone, and please let us know if you run into any issues! Stéfan

1 0

Announcement: scikit-image 0.13.0
by Juan Nunez-Iglesias March 29, 2017

March 29, 2017

We're happy to (finally) announce the release of scikit-image v0.13.0! Special thanks to our many contributors for making it possible! This release is the result of over a year of work, with over 200 pull requests by 82 contributors. Linux and macOS wheels are available now on PyPI <https://pypi.python.org/pypi/scikit-image>, together with a source distribution. Use "pip install -U scikit-image" to get the latest version! Packages on conda-forge, Windows wheels, and Debian packages should be available within the next few days. scikit-image is an image processing toolbox for SciPy that includes algorithms for segmentation, geometric transformations, color space manipulation, analysis, filtering, morphology, feature detection, and more. For more information, examples, and documentation, please visit our website: http://scikit-image.org and our gallery of examples http://scikit-image.org/docs/dev/auto_examples/ Highlights ---------- - Improved n-dimensional image support. This release adds nD support to: * ``regionprops`` computation for centroids (#2083) * ``segmentation.clear_border`` (#2087) * Hessian matrix (#2194) - In addition, the following new functions support nD images: * new wavelet denoising function, ``restoration.denoise_wavelet`` (#1833, #2190, #2238, #2240, #2241, #2242, #2462) * new thresholding functions, ``filters.threshold_sauvola`` and ``filters.threshold_niblack`` (#2266, #2441) * new local maximum, local minimum, hmaxima, hminima functions (#2449) - Grey level co-occurrence matrix (GLCM) now works with uint16 images - ``filters.try_all_threshold`` to rapidly see output of various thresholding methods - Frangi and Hessian filters (2D only) (#2153) - New *compact watershed* algorithm in ``segmentation.watershed`` (#2211) - New *shape index* algorithm in ``feature.shape_index`` (#2312) New functions and features -------------------------- - Add threshold minimum algorithm (#2104) - Implement mean and triangle thresholding (#2126) - Add Frangi and Hessian filters (#2153) - add bbox_area to region properties (#2187) - colorconv: Add rgba2rgb() (#2181) - Lewiner marching cubes algorithm (#2052) - image inversion (#2199) - wavelet denoising (from #1833) (#2190) - routine to estimate the noise standard deviation from an image (#1837) - Add compact watershed and clean up existing watershed (#2211) - Added the missing 'grey2rgb' function. (#2316) - Shape index (#2312) - Fundamental and essential matrix 8-point algorithm (#1357) - Add YUV, YIQ, YPbPr, YCbCr colorspaces - Detection of local extrema from morphology (#2449) - shannon entropy (#2416) Documentation improvements -------------------------- - add details about github SSH keys in contributing page (#2073) - Add example for felzenszwalb image segmentation (#2096) - Sphinx gallery for example gallery (#2078) - Improved region boundary RAG docs (#2106) - Add gallery Lucy-Richardson deconvolution algorithm (#2376) - Gallery: Use Horse to illustrate Convex Hull (#2431) - Add working with OpenCV in user guide (#2519) Code improvements ----------------- - Remove lena image from test suite (#1985) - Remove duplicate mean calculation in skimage.feature.match_template (#1980) - Add nD support to clear_border (#2087) - Add uint16 images support for co-occurrence matrix (#2095) - Add default parameters for Gaussian and median filters (#2151) - try_all to choose the best threshold algorithm (#2110) - Add support for multichannel in Felzenszwalb segmentation (#2134) - Improved SimilarityTransform, new EuclideanTransform class (#2044) - ENH: Speed up Hessian matrix computation (#2194) - add n-dimensional support to denoise_wavelet (#2242) - Speedup ``inpaint_biharmonic`` (#2234) - Update hessian matrix code to include order kwarg (#2327) - Handle cases for label2rgb where input labels are negative and/or nonconsecutive (#2370) - Added watershed_line parameter (#2393) API Changes ----------- - Remove deprecated ``filter`` module. Use ``filters`` instead. (#2023) - Remove ``skimage.filters.canny`` links. Use ``feature.canny`` instead. (#2024) - Removed Python 2.6 support and related checks (#2033) - Remove deprecated {h/v}sobel, {h/v}prewitt, {h/v}scharr, roberts_{positive/negative} filters (#2159) - Remove deprecated ``_mode_deprecations`` (#2156) - Remove deprecated None defaults in ``rescale_intensity`` (#2161) - Parameters ``ntiles_x`` and ``ntiles_y`` have been removed from ``exposure.equalize_adapthist`` - The minimum NumPy version is now 1.11, and the minimum SciPy version is now 0.17 Deprecations ------------ - clip_negative will be set to false by default in version 0.15 (func: dtype_limits) (#2228) - Deprecate "dynamic_range" in favor of "data_range" (#2384) - The default value of the ``circle`` argument to ``radon`` and ``iradon`` transforms will be ``True`` in 0.15 (#2235) - The default value of ``multichannel`` for ``denoise_bilateral`` and ``denoise_nl_means`` will be ``False`` in 0.15 - The default value of ``block_norm`` in ``feature.hog`` will be L2-Hysteresis in 0.15. - The ``threshold_adaptive`` function is deprecated. Use ``threshold_local`` instead. - The default value of ``mode`` in ``transform.swirl``, ``resize``, and ``rescale`` will be "reflect" in 0.15. For a complete list of contributors and pull requests merged in this release, please see our release notes online: https://github.com/scikit-image/scikit-image/blob/master/doc/release/releas… Please spread the word, including on Twitter <https://twitter.com/jnuneziglesias/status/847145183250665472>! Enjoy! Juan.

1 0

Announcement: scikit-image 0.13.0
by Juan Nunez-Iglesias March 29, 2017

March 29, 2017

We're happy to (finally) announce the release of scikit-image v0.13.0! Special thanks to all our contributors who made this possible. Linux and macOS wheels are available now on PyPI, as well as a source distribution. A conda-forge package, Windows wheels, and Debian packages should arrive in the coming days. scikit-image is an image processing toolbox for SciPy that includes algorithms for segmentation, geometric transformations, color space manipulation, analysis, filtering, morphology, feature detection, and more. For more information, examples, and documentation, please visit our website: http://scikit-image.org and our gallery of examples http://scikit-image.org/docs/dev/auto_examples/ Highlights ---------- This release is the result of a year of work, with over 200 pull requests by 82 contributors. Highlights include: - Improved n-dimensional image support. This release adds nD support to: * ``regionprops`` computation for centroids (#2083) * ``segmentation.clear_border`` (#2087) * Hessian matrix (#2194, #2327) - In addition, the following new functions support nD images: * new wavelet denoising function, ``restoration.denoise_wavelet`` (#1833, #2190, #2238, #2240, #2241, #2242, #2462) * new thresholding functions, ``filters.threshold_sauvola`` and ``filters.threshold_niblack`` (#2266, #2441) * new local maximum, local minimum, hmaxima, hminima functions (#2449) - Grey level co-occurrence matrix (GLCM) now works with uint16 images - ``filters.try_all_threshold`` to rapidly see output of various thresholding methods - Frangi and Hessian filters (2D only) (#2153) - New *compact watershed* algorithm in ``segmentation.watershed`` (#2211) - New *shape index* algorithm in ``feature.shape_index`` (#2312) New functions and features -------------------------- - Add threshold minimum algorithm (#2104) - Implement mean and triangle thresholding (#2126) - Add Frangi and Hessian filters (#2153) - add bbox_area to region properties (#2187) - colorconv: Add rgba2rgb() (#2181) - Lewiner marching cubes algorithm (#2052) - image inversion (#2199) - wavelet denoising (from #1833) (#2190) - routine to estimate the noise standard deviation from an image (#1837) - Add compact watershed and clean up existing watershed (#2211) - Added the missing 'grey2rgb' function. (#2316) - Shape index (#2312) - Fundamental and essential matrix 8-point algorithm (#1357) - Add YUV, YIQ, YPbPr, YCbCr colorspaces - Detection of local extrema from morphology (#2449) - shannon entropy (#2416) Documentation improvements -------------------------- - add details about github SSH keys in contributing page (#2073) - Add example for felzenszwalb image segmentation (#2096) - Sphinx gallery for example gallery (#2078) - Improved region boundary RAG docs (#2106) - Add gallery Lucy-Richardson deconvolution algorithm (#2376) - Gallery: Use Horse to illustrate Convex Hull (#2431) - Add working with OpenCV in user guide (#2519) Code improvements ----------------- - Remove lena image from test suite (#1985) - Remove duplicate mean calculation in skimage.feature.match_template (#1980) - Add nD support to clear_border (#2087) - Add uint16 images support for co-occurrence matrix (#2095) - Add default parameters for Gaussian and median filters (#2151) - try_all to choose the best threshold algorithm (#2110) - Add support for multichannel in Felzenszwalb segmentation (#2134) - Improved SimilarityTransform, new EuclideanTransform class (#2044) - ENH: Speed up Hessian matrix computation (#2194) - add n-dimensional support to denoise_wavelet (#2242) - Speedup ``inpaint_biharmonic`` (#2234) - Update hessian matrix code to include order kwarg (#2327) - Handle cases for label2rgb where input labels are negative and/or nonconsecutive (#2370) - Added watershed_line parameter (#2393) API Changes ----------- - Remove deprecated ``filter`` module. Use ``filters`` instead. (#2023) - Remove ``skimage.filters.canny`` links. Use ``feature.canny`` instead. (#2024) - Removed Python 2.6 support and related checks (#2033) - Remove deprecated {h/v}sobel, {h/v}prewitt, {h/v}scharr, roberts_{positive/negative} filters (#2159) - Remove deprecated ``_mode_deprecations`` (#2156) - Remove deprecated None defaults in ``rescale_intensity`` (#2161) - Parameters ``ntiles_x`` and ``ntiles_y`` have been removed from ``exposure.equalize_adapthist`` - The minimum NumPy version is now 1.11, and the minimum SciPy version is now 0.17 Deprecations ------------ - clip_negative will be set to false by default in version 0.15 (func: dtype_limits) (#2228) - Deprecate "dynamic_range" in favor of "data_range" (#2384) - The default value of the ``circle`` argument to ``radon`` and ``iradon`` transforms will be ``True`` in 0.15 (#2235) - The default value of ``multichannel`` for ``denoise_bilateral`` and ``denoise_nl_means`` will be ``False`` in 0.15 - The default value of ``block_norm`` in ``feature.hog`` will be L2-Hysteresis in 0.15. - The ``threshold_adaptive`` function is deprecated. Use ``threshold_local`` instead. - The default value of ``mode`` in ``transform.swirl``, ``resize``, and ``rescale`` will be "reflect" in 0.15. For a complete list of contributors to this release, and PRs merged, please see the online release notes: https://github.com/scikit-image/scikit-image/blob/master/doc/release/releas… Enjoy!

1 0

Sprint at SciPy2017
by Stefan van der Walt March 29, 2017

March 29, 2017

Hi, everyone If you are interested in hosting a sprint with me at SciPy2017, please get in touch directly. Thanks! Stéfan

1 0

Re: [scikit-image] color.lab2rgb output differs from colormine.org
by Juan Nunez-Iglesias March 27, 2017

March 27, 2017

Hi Michael, Yes, it’s common in Python and you should generally work in floating point until the very final step of processing. This will help avoid accumulation of rounding errors. See our documentation on this here: http://scikit-image.org/docs/dev/user_guide/data_types.html Happy coding! Juan. On 27 Mar 2017, 10:46 AM -0400, Michael O'Brien <mob.irl(a)gmail.com>, wrote: > Using http://www.easyrgb.com/index.php > > CIE-L*ab = 48.000 -14.000 25.000 > > equals > > RGB 0-255 = 105.30 119.53 70.87 > RGB 0-0.1 = 0.41293 0.46875 0.27791 > > XYZ = 13.654 16.794 8.461 > > Illuminant = D65 > Observer = 2° (1931) > > So Ryan was correct to say the values are scaled between 0 and 1 (I was using the wrong row in my re-conversion to 0 to 255 scale) > This is the 1st time I've come across RGB scaled to 0 and 1. Is its common in python and does the other scikit-image functions handle the 0 -1 scaled values or should I work with 0 to 255 values to be safe? > > > > > > On Mon, 27 Mar 2017 at 14:36 Michael O'Brien <mob.irl(a)gmail.com> wrote: > > > I'm not sure what illuminant colormine uses but if you multiple by 255 you get 151,154,141 which it getting closer to what I expect (greenish colour) with 105,119,70 still being much closer > > > > > > On Mon, 27 Mar 2017, 14:22 Ryan Anderson, <randerson(a)resonon.com> wrote: > > > > It looks to me as if lab2rgb is returning colors scaled as floats between 0 and 1 while colormine.org is returning 8 bit integers scaled 0-255. > > > > > > > > Try multiplying by 255 and comparing the outputs. > > > > > > > > On Mar 27, 2017 7:02 AM, "Michael O'Brien" <mob.irl(a)gmail.com> wrote: > > > > > > > > > > > > > > > Hi all, > > > > > > > > > > > > > > > If I use the following array of LAB values > > > > > array([[[ 48.07851346, -14.69400135, 25.2170507 ], > > > > > [ 63.21232753, -3.65444198, 6.67734193], > > > > > [ 45.80027308, -2.47952304, 11.5073341 ], > > > > > [ 46.84900559, -10.37025167, 19.92527996], > > > > > [ 44.92476952, 19.36861848, 2.97238555], > > > > > [ 35.50132805, 5.89086564, 1.95259827]]]) > > > > > and use the following code to pass the array to lab2rgb > > > > > rgbcolours =color.lab2rgb(ClusterColours.values[np.newaxis]) > > > > > I get the array > > > > > array([[[ 0.40929448, 0.47071505, 0.27701891], > > > > > [ 0.59383913, 0.60611158, 0.55329837], > > > > > [ 0.4393785 , 0.4276561 , 0.34999225], > > > > > [ 0.4159481 , 0.4516056 , 0.3026519 ], > > > > > [ 0.54449997, 0.36963636, 0.4001209 ], > > > > > [ 0.36970012, 0.3145826 , 0.315974 ]]]) > > > > > but on colormine.org if I enter the 1st of the LAB values I get 105,119,70 for RGB and not colours so close to black > > > > > > > > > > Where am I going wrong? > > > > > > > > > > _______________________________________________ > > > > > scikit-image mailing list > > > > > scikit-image(a)python.org > > > > > https://mail.python.org/mailman/listinfo/scikit-image > > > > > > _______________________________________________ > scikit-image mailing list > scikit-image(a)python.org > https://mail.python.org/mailman/listinfo/scikit-image

1 0

Re: [scikit-image] color.lab2rgb output differs from colormine.org
by Michael O'Brien March 27, 2017

March 27, 2017

Using http://www.easyrgb.com/index.php *CIE-L*ab = 48.000 -14.000 25.000* *equals* *RGB 0-255 = 105.30 119.53 70.87* *RGB 0-0.1 = 0.41293 0.46875 0.27791* XYZ = 13.654 16.794 8.461 Illuminant = D65 Observer = 2° (1931) So Ryan was correct to say the values are scaled between 0 and 1 (I was using the wrong row in my re-conversion to 0 to 255 scale) This is the 1st time I've come across RGB scaled to 0 and 1. Is its common in python and does the other scikit-image functions handle the 0 -1 scaled values or should I work with 0 to 255 values to be safe? On Mon, 27 Mar 2017 at 14:36 Michael O'Brien <mob.irl(a)gmail.com> wrote: > I'm not sure what illuminant colormine uses but if you multiple by 255 you > get 151,154,141 which it getting closer to what I expect (greenish colour) > with 105,119,70 still being much closer > > On Mon, 27 Mar 2017, 14:22 Ryan Anderson, <randerson(a)resonon.com> wrote: > > It looks to me as if lab2rgb is returning colors scaled as floats between > 0 and 1 while colormine.org is returning 8 bit integers scaled 0-255. > > Try multiplying by 255 and comparing the outputs. > > On Mar 27, 2017 7:02 AM, "Michael O'Brien" <mob.irl(a)gmail.com> wrote: > > Hi all, > > > If I use the following array of LAB values > array([[[ 48.07851346, -14.69400135, 25.2170507 ], > [ 63.21232753, -3.65444198, 6.67734193], > [ 45.80027308, -2.47952304, 11.5073341 ], > [ 46.84900559, -10.37025167, 19.92527996], > [ 44.92476952, 19.36861848, 2.97238555], > [ 35.50132805, 5.89086564, 1.95259827]]]) > and use the following code to pass the array to lab2rgb > rgbcolours =color.lab2rgb(ClusterColours.values[np.newaxis]) > I get the array > array([[[ 0.40929448, 0.47071505, 0.27701891], > [ 0.59383913, 0.60611158, 0.55329837], > [ 0.4393785 , 0.4276561 , 0.34999225], > [ 0.4159481 , 0.4516056 , 0.3026519 ], > [ 0.54449997, 0.36963636, 0.4001209 ], > [ 0.36970012, 0.3145826 , 0.315974 ]]]) > but on colormine.org if I enter the 1st of the LAB values I get 105,119,70 for RGB and not colours so close to black > > Where am I going wrong? > > > _______________________________________________ > scikit-image mailing list > scikit-image(a)python.org > https://mail.python.org/mailman/listinfo/scikit-image > >

1 0

Re: [scikit-image] color.lab2rgb output differs from colormine.org
by Michael O'Brien March 27, 2017

March 27, 2017

I'm not sure what illuminant colormine uses but if you multiple by 255 you get 151,154,141 which it getting closer to what I expect (greenish colour) with 105,119,70 still being much closer On Mon, 27 Mar 2017, 14:22 Ryan Anderson, <randerson(a)resonon.com> wrote: > It looks to me as if lab2rgb is returning colors scaled as floats between > 0 and 1 while colormine.org is returning 8 bit integers scaled 0-255. > > Try multiplying by 255 and comparing the outputs. > > On Mar 27, 2017 7:02 AM, "Michael O'Brien" <mob.irl(a)gmail.com> wrote: > > Hi all, > > > If I use the following array of LAB values > array([[[ 48.07851346, -14.69400135, 25.2170507 ], > [ 63.21232753, -3.65444198, 6.67734193], > [ 45.80027308, -2.47952304, 11.5073341 ], > [ 46.84900559, -10.37025167, 19.92527996], > [ 44.92476952, 19.36861848, 2.97238555], > [ 35.50132805, 5.89086564, 1.95259827]]]) > and use the following code to pass the array to lab2rgb > rgbcolours =color.lab2rgb(ClusterColours.values[np.newaxis]) > I get the array > array([[[ 0.40929448, 0.47071505, 0.27701891], > [ 0.59383913, 0.60611158, 0.55329837], > [ 0.4393785 , 0.4276561 , 0.34999225], > [ 0.4159481 , 0.4516056 , 0.3026519 ], > [ 0.54449997, 0.36963636, 0.4001209 ], > [ 0.36970012, 0.3145826 , 0.315974 ]]]) > but on colormine.org if I enter the 1st of the LAB values I get 105,119,70 for RGB and not colours so close to black > > Where am I going wrong? > > > _______________________________________________ > scikit-image mailing list > scikit-image(a)python.org > https://mail.python.org/mailman/listinfo/scikit-image > >

1 0

Re: [scikit-image] color.lab2rgb output differs from colormine.org
by Ryan Anderson March 27, 2017

March 27, 2017

It looks to me as if lab2rgb is returning colors scaled as floats between 0 and 1 while colormine.org is returning 8 bit integers scaled 0-255. Try multiplying by 255 and comparing the outputs. On Mar 27, 2017 7:02 AM, "Michael O'Brien" <mob.irl(a)gmail.com> wrote: Hi all, If I use the following array of LAB values array([[[ 48.07851346, -14.69400135, 25.2170507 ], [ 63.21232753, -3.65444198, 6.67734193], [ 45.80027308, -2.47952304, 11.5073341 ], [ 46.84900559, -10.37025167, 19.92527996], [ 44.92476952, 19.36861848, 2.97238555], [ 35.50132805, 5.89086564, 1.95259827]]]) and use the following code to pass the array to lab2rgb rgbcolours =color.lab2rgb(ClusterColours.values[np.newaxis]) I get the array array([[[ 0.40929448, 0.47071505, 0.27701891], [ 0.59383913, 0.60611158, 0.55329837], [ 0.4393785 , 0.4276561 , 0.34999225], [ 0.4159481 , 0.4516056 , 0.3026519 ], [ 0.54449997, 0.36963636, 0.4001209 ], [ 0.36970012, 0.3145826 , 0.315974 ]]]) but on colormine.org if I enter the 1st of the LAB values I get 105,119,70 for RGB and not colours so close to black Where am I going wrong? _______________________________________________ scikit-image mailing list scikit-image(a)python.org https://mail.python.org/mailman/listinfo/scikit-image

1 0

Re: [scikit-image] color.lab2rgb output differs from colormine.org
by Isaac Gerg March 27, 2017

March 27, 2017

What illuminant standard is colormine.org using? Looks like scikit-image is using D65. https://github.com/scikit-image/scikit-image/blob/1c4646c25c9e1158bf7134973… On Mon, Mar 27, 2017 at 9:01 AM, Michael O'Brien <mob.irl(a)gmail.com> wrote: > Hi all, > > > If I use the following array of LAB values > array([[[ 48.07851346, -14.69400135, 25.2170507 ], > [ 63.21232753, -3.65444198, 6.67734193], > [ 45.80027308, -2.47952304, 11.5073341 ], > [ 46.84900559, -10.37025167, 19.92527996], > [ 44.92476952, 19.36861848, 2.97238555], > [ 35.50132805, 5.89086564, 1.95259827]]]) > and use the following code to pass the array to lab2rgb > rgbcolours =color.lab2rgb(ClusterColours.values[np.newaxis]) > I get the array > array([[[ 0.40929448, 0.47071505, 0.27701891], > [ 0.59383913, 0.60611158, 0.55329837], > [ 0.4393785 , 0.4276561 , 0.34999225], > [ 0.4159481 , 0.4516056 , 0.3026519 ], > [ 0.54449997, 0.36963636, 0.4001209 ], > [ 0.36970012, 0.3145826 , 0.315974 ]]]) > but on colormine.org if I enter the 1st of the LAB values I get 105,119,70 for RGB and not colours so close to black > > Where am I going wrong? > > > _______________________________________________ > scikit-image mailing list > scikit-image(a)python.org > https://mail.python.org/mailman/listinfo/scikit-image > >

1 0