Mailman 3 June 2017 - scikit-image

Re: [scikit-image] Why does img_as force incoming data
by Cedric Espenel 21 Jun '17

21 Jun '17

Hi Dave, I think this scikit-image should answers your questions: http://scikit-image.org/docs/stable/user_guide/data_types.html Best, Cedric On Wed, Jun 21, 2017 at 4:30 PM David Protter <david.protter(a)gmail.com> wrote: > Thank you! Will this also work going back the other way? (Float to uint?) > > On Jun 21, 2017 5:16 PM, "Stefan van der Walt" <stefanv(a)berkeley.edu> > wrote: > >> >> >> >> >> >> >> >> >> >> >> Hi Dave >> >> >> >> >> >> On Wed, Jun 21, 2017, at 15:22, David Protter wrote: >> >> >> Hi all, new here and having a lot of fun using Skimage for scientific >> image analysis. >> >> >> >> >> >> >> >> Glad to hear it! >> >> >> >> >> >> I’m doign some normalization on images coming in as uint16, to expand >> their dynamic range. After normalization, images are float64, and I’m >> trying to convert them back to uint16. However, it seems like all the >> img_as functions ignore the kind of float coming in, since anything of kind >> float is constrained to [-1 to 1]. Below is the section out of the >> convert() function that seems to do this. >> >> >> >> >> >> >> >> Floating point images are expected to be between -1 and 1, as described >> here: >> >> >> >> >> >> http://scikit-image.org/docs/stable/user_guide/data_types.html >> >> >> >> >> >> However, many functions support working on data in its original range, by >> specifying `preserve_range=True`. >> >> >> >> >> >> To convert your data from uint16 to float without changing its range, do: >> >> >> >> >> >> img_float = img.astype(float) >> >> >> >> >> >> Best regards >> >> >> Stéfan >> >> >> >> >> >> >> >> >> >> > > _______________________________________________ > > scikit-image mailing list > > scikit-image(a)python.org > > https://mail.python.org/mailman/listinfo/scikit-image > >

1 0

Re: [scikit-image] Why does img_as force incoming data
by David Protter 21 Jun '17

21 Jun '17

Thank you! Will this also work going back the other way? (Float to uint?) On Jun 21, 2017 5:16 PM, "Stefan van der Walt" <stefanv(a)berkeley.edu> wrote: > Hi Dave > > On Wed, Jun 21, 2017, at 15:22, David Protter wrote: > > Hi all, new here and having a lot of fun using Skimage for scientific > image analysis. > > > Glad to hear it! > > I’m doign some normalization on images coming in as uint16, to expand > their dynamic range. After normalization, images are float64, and I’m > trying to convert them back to uint16. However, it seems like all the > img_as functions ignore the kind of float coming in, since anything of kind > float is constrained to [-1 to 1]. Below is the section out of the > convert() function that seems to do this. > > > Floating point images are expected to be between -1 and 1, as described > here: > > http://scikit-image.org/docs/stable/user_guide/data_types.html > > However, many functions support working on data in its original range, by > specifying `preserve_range=True`. > > To convert your data from uint16 to float without changing its range, do: > > img_float = img.astype(float) > > Best regards > Stéfan > >

1 0

Re: [scikit-image] Why does img_as force incoming data
by Stefan van der Walt 21 Jun '17

21 Jun '17

Hi Dave On Wed, Jun 21, 2017, at 15:22, David Protter wrote: > Hi all, new here and having a lot of fun using Skimage for scientific > image analysis. Glad to hear it! > I’m doign some normalization on images coming in as uint16, to expand > their dynamic range. After normalization, images are float64, and I’m > trying to convert them back to uint16. However, it seems like all the > img_as functions ignore the kind of float coming in, since anything of > kind float is constrained to [-1 to 1]. Below is the section out of > the convert() function that seems to do this. Floating point images are expected to be between -1 and 1, as described here: http://scikit-image.org/docs/stable/user_guide/data_types.html However, many functions support working on data in its original range, by specifying `preserve_range=True`. To convert your data from uint16 to float without changing its range, do: img_float = img.astype(float) Best regards Stéfan

1 0

Why does img_as force incoming data
by David Protter 21 Jun '17

21 Jun '17

Hi all, new here and having a lot of fun using Skimage for scientific image analysis. I’m doign some normalization on images coming in as uint16, to expand their dynamic range. After normalization, images are float64, and I’m trying to convert them back to uint16. However, it seems like all the img_as functions ignore the kind of float coming in, since anything of kind float is constrained to [-1 to 1]. Below is the section out of the convert() function that seems to do this. # float -> any if kind_in == 'f': if np.min(image) < -1.0 or np.max(image) > 1.0: raise ValueError("Images of type float must be between -1 and 1.") if kind_out == 'f': # float -> float if itemsize_in > itemsize_out: prec_loss() return image.astype(dtype_out) Obviously my float64 array returns kind_in as ‘f’, but has values outside of this range. I’m pretty new to all this, am I missing something obvious? Thanks! Dave

1 0

回复:Re: Fitting Ellipse Model
by imagepy＠sina.com 13 Jun '17

13 Jun '17

but i think the opencv's hu moment is a vector implement(integrate contours), but scikit-image one is raster implement(statistic the pixcel coordinate) -------------------------------- 来自我的新浪邮箱android客户端 -------- 原始邮件 -------- 发件人:Stefan van der Walt 时间:2017年6月13日 08:05(星期二) 收件人:Surya Kasturi Mailing list for scikit-image (http://scikit-image.org) 主题:Re: [scikit-image] Fitting Ellipse Model Hi Surya On Mon, Jun 12, 2017, at 16:54, Surya Kasturi wrote: I’ve commented this snippet on Github. Anyway, is there any functionality similar to matchShapes provided in OpenCV? We have Hu moments implemented, so you can plug those directly into the formulas shown at the link you gave. Best regards Stéfan _______________________________________________ scikit-image mailing list scikit-image(a)python.org https://mail.python.org/mailman/listinfo/scikit-image

2 1

Re: [scikit-image] Fitting Ellipse Model
by Surya Kasturi 12 Jun '17

12 Jun '17

Hi Stefan, I’ve commented this snippet on Github. Anyway, is there any functionality similar to matchShapes <http://docs.opencv.org/2.4/modules/imgproc/doc/structural_analysis_and_shap…> provided in OpenCV? Thanks Surya > On Jun 12, 2017, at 2:23 PM, Stefan van der Walt <stefanv(a)berkeley.edu> wrote: > > Hi Surya > > On Sun, Jun 11, 2017, at 20:59, Surya Kasturi wrote: >> The below is output. Shouldn’t both ellipses in the image overlap since I’m using data points of a perfect ellipse to estimate a new ellipse? > > I believe you've uncovered a bug. Ellipse fitting and ellipse drawing seem to use different coordinate conventions. Please continue the discussion here: > > https://github.com/scikit-image/scikit-image/issues/2646 <https://github.com/scikit-image/scikit-image/issues/2646> > > Best regards > Stéfan >

2 1

Fitting Ellipse Model
by Surya Kasturi 12 Jun '17

12 Jun '17

Hi, I am trying to understand how measure.EllipseModel is working. Following is the snippet. x1, y1, a1, b1 = 3, 3, 20, 40 rr1, cc1 = draw.ellipse_perimeter(x1, y1, a1, b1) # some ellipse # Estimate ellipse model with the above points model = EllipseModel() model.estimate(np.array([rr1,cc1]).T) x2, y2, a2, b2, theta = np.array(model.params, dtype=int) rr2, cc2 = draw.ellipse_perimeter(x2, y2, a2, b2, orientation=theta) # estimated ellipse ax.scatter(rr1, cc1, s=5, label="actual") ax.scatter(rr2, cc2, s=5, label="estimated") The below is output. Shouldn’t both ellipses in the image overlap since I’m using data points of a perfect ellipse to estimate a new ellipse? Thanks Surya

2 1