XKCD style graphs?
Locked
12
12
 
|
|
Hi Fernando,
Le 04/10/2012 09:16, Fernando Perez a écrit : > This would make for an awesome couple of examples for the gallery, the > mathematica solutions look really pretty cool: > > http://mathematica.stackexchange.com/questions/11350/xkcd-style-graphs I've never used Mathematica so that it's pretty difficult for me to understand the following lines of code which I guess do the main job of distorting the image xkcdDistort[p_] := Module[{r, ix, iy}, r = ImagePad[Rasterize@p, 10, Padding -> White]; {ix, iy} = Table[RandomImage[{-1, 1}, ImageDimensions@r]~ImageConvolve~ GaussianMatrix[10], {2}]; ImagePad[ImageTransformation[r, # + 15 {ImageValue[ix, #], ImageValue[iy, #]} &, DataRange -> Full], -5]]; Is there somebody there that can describe this algorithm with words (English or Python ;-)) ? I feel like the key point is about adressing the rasterized plot image "r" with some slightly randomized indices "ix" and "iy". However, I really don't get the step that generates these indices. Best, Pierre ------------------------------------------------------------------------------ Don't let slow site performance ruin your business. Deploy New Relic APM Deploy New Relic app performance management and know exactly what is happening inside your Ruby, Python, PHP, Java, and .NET app Try New Relic at no cost today and get our sweet Data Nerd shirt too! http://p.sf.net/sfu/newrelic-dev2dev _______________________________________________ Matplotlib-users mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/matplotlib-users |
|
|
|
|
|
|
|
|
In reply to this post by Phil Elson-2
Le 04/10/2012 14:29, Phil Elson a écrit :
> Damon, I love the solution! I do wonder whether we could do some > quirky transform on the lines to achieve a similar result, rather than > manipulating the data before plotting it. The benefit is that > everything should then get randomly Xkcd-ed automatically - maybe I > will save that one for a rainy day.... > > A different solution to get the shaken effect on every graphic items is the post-processing of a raster rendering of the plot. I think this is what was proposed with Mathematica though I'm really unfamiliar with its syntax One way I see to "shake" on image would be to use scipy.ndimage.interpolation.map_coordinates [1] to interpolate the rastered plot image with a "shaken grid". This shaken grid would be a regular 2D indexing grid + some 2D noise, carefully tuned to have a bit of spatial correlation. I'm not so familiar with image processing in Python though, so there may be better solutions I'm not aware of. Best, Pierre [1] http://docs.scipy.org/doc/scipy/reference/generated/scipy.ndimage.interpolation.map_coordinates.htm ------------------------------------------------------------------------------ Don't let slow site performance ruin your business. Deploy New Relic APM Deploy New Relic app performance management and know exactly what is happening inside your Ruby, Python, PHP, Java, and .NET app Try New Relic at no cost today and get our sweet Data Nerd shirt too! http://p.sf.net/sfu/newrelic-dev2dev _______________________________________________ Matplotlib-users mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/matplotlib-users |
|
|
|
|
|
In reply to this post by Jason Grout-5
Le 04/10/2012 16:03, Jason Grout a écrit :
> f@r means f(r) > > a~ImageConvolve~b means ImageConvolve(a,b) (~ treats an operator as infix) > > Table[..., {2}] means [... for i in range(2)] > > #+1& is a lambda function lambda x: x+1 > > So I think it goes something like: > > def xkcdDistort(p): > r = ImagePad(Rasterize(p), 10, Padding='White') > (ix, iy) = [ImageConvolve(RandomImage([-1,1], ImageDimensions(r)), > GaussianMatrix(10)) > for i in range(2)] > return ImagePad(ImageTransformation(r, > lambda coord: (coord[0]+15*ImageValue(ix, coord), > coord[1]+15*ImageValue(iy, coord)), > DataRange='Full'), > -5) It's the first time I encounter Mathematica syntax. Some of these functional notations are not so easy to follow for my unexperienced eyes but it makes this Mathematica code nicely compact. So I think this code indeed resamples the rastered plot image on a shaken coordinate grid. I kind of understand that the noise on coordinates is spatially smoothed by a 10px Gaussian Point Spread Function (if I understand correctly...) Best, Pierre ------------------------------------------------------------------------------ Don't let slow site performance ruin your business. Deploy New Relic APM Deploy New Relic app performance management and know exactly what is happening inside your Ruby, Python, PHP, Java, and .NET app Try New Relic at no cost today and get our sweet Data Nerd shirt too! http://p.sf.net/sfu/newrelic-dev2dev _______________________________________________ Matplotlib-users mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/matplotlib-users |
|
|
In reply to this post by Michael Droettboom-3
Le 04/10/2012 16:11, Michael Droettboom a écrit :
> Yes -- this would be a great application for the path filtering > infrastructure that matplotlib has. Sounds way cooler than post-processing a raster plot image ! I'm not aware of this path filtering infrastructure. I guess it's a deeply buried facility which is not accessible in the "Python user space" ? Best, Pierre ------------------------------------------------------------------------------ Don't let slow site performance ruin your business. Deploy New Relic APM Deploy New Relic app performance management and know exactly what is happening inside your Ruby, Python, PHP, Java, and .NET app Try New Relic at no cost today and get our sweet Data Nerd shirt too! http://p.sf.net/sfu/newrelic-dev2dev _______________________________________________ Matplotlib-users mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/matplotlib-users |
|
|
|
|
|
|
|
|
|
|
|
|
In reply to this post by Damon McDougall
Le 04/10/2012 16:54, Damon McDougall a écrit :
> Adding Gaussian noise to each point on a function doesn't look nice. > That's why I produced a random function in Fourier space first. That > way, random functions still have some sense of smoothness. Mathematica code seems to use a Gaussian *smoothing* of a uniform noise. I understand this as the spatial-domain-way (using convolution) to get some smoothness while you've taken the frequency-domain path. It's a matter of taste and I guess that both ways should be ok ! Best, Pierre ------------------------------------------------------------------------------ Don't let slow site performance ruin your business. Deploy New Relic APM Deploy New Relic app performance management and know exactly what is happening inside your Ruby, Python, PHP, Java, and .NET app Try New Relic at no cost today and get our sweet Data Nerd shirt too! http://p.sf.net/sfu/newrelic-dev2dev _______________________________________________ Matplotlib-users mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/matplotlib-users |
|
|
In reply to this post by Pierre Haessig
Le 04/10/2012 16:35, Pierre Haessig a écrit :
> So I think this code indeed resamples the rastered plot image on a > shaken coordinate grid. I kind of understand that the noise on > coordinates is spatially smoothed by a 10px Gaussian Point Spread > Function (if I understand correctly...) I've implemented this processing in a tiny "image_shake" script. https://gist.github.com/3834536 A nice occasion to learn how to use some scipy image processing functions... I've attached the before/after images because I didn't manage to put them in the Gist (it's not a plot image but gives the idea of line shaking). Now, I think it's unfortunately outside the frame of Fernando's challenge, because this script uses zero matplotlib methods!! Best, Pierre ------------------------------------------------------------------------------ Don't let slow site performance ruin your business. Deploy New Relic APM Deploy New Relic app performance management and know exactly what is happening inside your Ruby, Python, PHP, Java, and .NET app Try New Relic at no cost today and get our sweet Data Nerd shirt too! http://p.sf.net/sfu/newrelic-dev2dev _______________________________________________ Matplotlib-users mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/matplotlib-users  grid_image.png (4K) Download Attachment grid_image-shaken.png (29K) Download Attachment signature.asc (918 bytes) Download Attachment |
|
Loading...
| Powered by Nabble | Edit this page |