I'm having an issue with attempting to save some plots with transparent ellipsoids on them if I attempt to save them with .ps/.eps extensions.
Here's the plot saved as a .png:
If I choose to save it as a .ps/.eps here is what it looks like:
How I got around this, was to use ImageMagick to convert the original png to a ps. The only problem is that the image in png format is about 90k, and it becomes just under 4M after conversion. This is not good since I have a lot of these images, and it will take too much time to compile my latex document. Does anyone have a solution to this?
The problem is that eps does not support transparencies natively.
There are few options:
rasterize the image and embed in a eps file (like #Molly suggests) or exporting to pdf and converting with some external tool (like gs) (which usually relies as well on rasterization)
'mimic' transparency, giving a colour that looks like the transparent one on a given background.
I discussed this for sure once on the matplotlib mailing list, and I got the suggestion to rasterize, which is not feasible as you get either pixellized or huge figures. And they don't scale very nicely when put into, e.g., a publication.
I personally use the second approach, and although not ideal, I found it good enough. I wrote a small python script that implements the algorithm from this SO post to obtain a solid RGB representation of a colour with a give transparency
EDIT
In the specific case of your plot try to use the zorder keyword to order the parts plotted. Try to use zorder=10 for the blue ellipse, zorder=11 for the green and zorder=12 for the hexbins.
This way the blue should be below everything, then the green ellipse and finally the hexbins. And the plot should be readable also with solid colors. And if you like the shades of blue and green that you have in png, you can try to play with mimic_alpha.py.
EDIT 2
If you are 100% sure that you have to use eps, there are a couple of workarounds that come to my mind (and that are definitely uglier than your plot):
Just draw the ellipse borders on top of the hexbins.
Get centre and amplitude of each hexagon, (possibly discard all zero bins) and make a scatter plot using the same colour map as in hexbin and adjusting the marker size and shape as you like. You might want to redraw the ellipses borders on top of that
Another alternative would be to save them to pdf
savefig('myfigure.pdf')
That works with pdflatex, if that was the reason why you needed to use eps and not svg.
You can rasterize the figure before saving it to preserve transparency in the eps file:
ax.set_rasterized(True)
plt.savefig('rasterized_fig.eps')
I had the same problem. To avoid rasterizing, you can save the image as a pdf and then run (on unixish systems at least) in a terminal:
pdftops -eps my.pdf my.eps
Which gives a .eps file as output.
I solved this by:
1) adding a set_rasterization_zorder(1) when defining the figure area:
fxsize=16
fysize=8
f = figure(num=None, figsize=(fxsize, fysize), dpi=180, facecolor='w',
edgecolor='k')
plt.subplots_adjust(
left = (18/25.4)/fxsize,
bottom = (13/25.4)/fysize,
right = 1 - (8/25.4)/fxsize,
top = 1 - (8/25.4)/fysize)
subplots_adjust(hspace=0,wspace=0.1)
#f.suptitle('An overall title', size=20)
gs0 = gridspec.GridSpec(1, 2)
gs11 = gridspec.GridSpecFromSubplotSpec(1, 1, subplot_spec=gs0[0])
ax110 = plt.Subplot(f, gs11[0,0])
f.add_subplot(ax110)
ax110.set_rasterization_zorder(1)
2) a zorder=0 in each alpha=anynumber in the plot:
ax110.scatter(xs1,ys1 , marker='o', color='gray' , s=1.5,zorder=0,alpha=0.3)#, label=label_bg)
and
3) finally a rasterized=True when saving:
P.savefig(str(PLOTFILENAME)+'.eps', rasterized=True)
Note that this may not work as expected with the transparent keyword to savefig because an RGBA colour with alpha<1 on transparent background will be rendered the same as the RGB colour with alpha=1.
As mentioned above, the best and easiest choice (if you do not want to loose resolution) is to rasterized the figure
f = plt.figure()
f.set_rasterized(True)
ax = f.add_subplot(111)
ax.set_rasterized(True)
f.savefig('figure_name.eps',rasterized=True,dpi=300)
This way, you can manage the size by dpi option as well. In fact, you can also play with the zorder below you want to apply the rasterization:
ax.set_rasterization_zorder(0)
Note: It is important to keep f.set_rasterized(True) when you use plt.subplot and plt.subplot2grid functions. Otherwise, label and tick area will not appear in the .eps file
My solution is to export the plot as .eps, load it up to Inkscape for example, then Ungroup the plot, select the object that I want to set the transparency and just edit the Opacity of the Fill in the "Fill and Stroke" tab.
You can save the file as .svg if you want to tweak it later, or export the image for a publication.
If you are writing the academic paper in latex, I would recommend you export the .pdf file rather than .eps. The .pdf format supports transparency perfectly and has good compression efficiency, and most importantly, can be easily edited in Adobe Illustrator.
If you wanna further edit the graph (NOT EDITING DATA! I MEAN, FOR GOOD-LOOKING), you could open the exported graph, in Adobe Acrobat - Edit - Copy elements into Adobe Illustrator. The two software can handle everything perfectly.
I work happily with this method. Everything clear, editable and small-size. Hope can help.
Related
EDIT 2: Analyzed files.
After looking better at the gradient files I see how input/output signals of alpha levels are converted. This image shows how input alpha level (RED) is converted to output (BLUE), which looks clearly wrong and maybe is a bug. Or, maybe I'm wrong altogether and I'm doing something wrong.
EDIT: Added files.
The original is quite messy with 40+ layers. I've reproduced the problem with a single layer in a very simple case (see files below).
The commands used are:
convert gradient.psd[0] gradient-0.png
convert gradient.psd[1] gradient-1.png
It looks like the [0] and [1] semantic is not clear to me. I thought that [0] is how to merge all layers, while [1] is just the first layer.
NOTICE in gradient-0.png how conversion messed up the gradient!
These are the files:
gradient.psd
gradient-0.png
gradient-1.png
The image to the left is a PSD with multiple layers, one of which has a fading BLACK alpha channel ("shadow").
By painting on a layer ("backdrop") beneath the "shadow" layer, the colors are correctly darkened by it as expected.
What I do now, is convert with ImageMagick the PSD file into a PNG file, which does not have layers but an alpha channel. I expected the converter to preserve alpha channel data and, in fact, it does. But it somehow messes up the color of the pixels which become GRAY (or white?) instead of black.
The image to the right shows this effect. That is the resulting PNG file with a non-transparent backdrop to better show the undesired gray coloring.
The conversion command I use is as simple as it gets:
convert temp.psd[0] temp.png
Can someone explain what is happening? Or, better, help with converting it correctly?
I cannot use the GUI commands, this actually is an asset pipeline that will pump volumes of images like these hundreds of time, so scripting is absolutely necessary.
I am creating a GUI containing an image using the following code:
try
Imagenamehere = imread('Imagenamehere.jpg');
axes(handles.Logo)
image(Imagenamehere)
set(gca,'xtick',[],'ytick',[])
catch
msgbox('Please download all contents from the zipped file into working directory.')
end
The image shows up but for some reason is completely coloured blue as if put through a blue filter. I don't think it would be wise to upload the image but it is a simple logo coloured black and white.
Anyone know what could be causing this?
Check the size, type (probably uint8) and range of your image. It sounds like for some reason your images are being displayed with colormap as jet (the default), and possibly also that your range is not what MATLAB expects (e.g. 0 to 1 not 0 to 255), resulting in all your values being relatively low (blue on the jet colormap).
"black and white" is just one way of interpreting an image file which contains only two colors. MATLAB makes several assumptions when you pass data into a display function like image. If you don't specify colormap and image data range, it will make a guess based off things like data type.
One possibility is that your logo file is an indexed image. In these cases you need to do:
[Imagenamehere map] = imread('Imagenamehere.jpg');
colormap(map);
I'm using the print function in MATLAB to write images of plots, something like that
print(figure(1),'-dpng','-r300',filename);
But apparently the images are not overwritten, and the original images stay. I was using saveas before, which seems to overwrite the images, but print gives me more output options. Any ideas?
UPDATE: I ended up deleting the files before the printing with a different function.
You can use this:
im = frame2im(getframe(gcf,rec)); %Grabs image of plot as an image
imsave(im, filename); %save image
That syntax may not be 100%, its a while since I've used it.
Also be aware that this isn't perfect - I remember having issues with it grabbing a grey border around the edge of the plot. Also, I think the image may be based on a matlab screenshot.... just something to be aware of
Saving figures in matlab is rather troublesome, especially if the saved image should look like the original figure.
For myself i found the solution in using export_fig.
It's one of the most downloaded fileexchange files - maybe you should give it a try:
http://www.mathworks.de/matlabcentral/fileexchange/23629-export-fig
A small introduction to export_fig can be found at:
https://github.com/ojwoodford/export_fig/blob/master/README.md
I'm currently creating my figures in matlab to embed themvia latex into a pdf for later printing. I save the figures and save them via the script export_fig! Now I wonder which is the best way to go:
Which size of the matlab figure window to chose
Which -m option to take for the script? It will change the resolution and the size of the image...
I'm wondering about those points in regards to the following two points:
When chosing the figure-size bigger, there are more tickmarks shown and the single point markers are better visible
When using a small figure and using a big -m option, I still have only some tickmarks
When I generate a image which is quite huge (e.g. resolution 300 and still 2000*2000px) and than embed it into the document: Does this than look ugly? Will this be embedded in a nice scaling mode or is it the same ugliness as if you upload a 1000*1000px image onto a homepage and embed it via the widht and height tags in html -> the browser displays it quite ugly because the browser doesn't do a real resize. So it looks unsharp and ugly.
Thanks in advance!
The MATLAB plots are internally described as vector graphics, and PDF files are also described using vector graphics. Rendering the plot to a raster format is a bad idea, because you end up having to choose resolution and end up with bigger files.
Just save the plot to EPS format, which can be directly embedded into a PDF file using latex. I usually save my MATLAB plots for publication using:
saveas(gcf, 'plot.eps', 'epsc');
and embed them directly into my latex file using:
\includegraphics[width=0.7\linewidth]{plot.eps}
Then, you only need to choose the proportion of the line the image is to take (in this case, 70%).
Edit: IrfanView and others (XnView) don't display EPS very well. You can open them in Adobe Illustrator to get a better preview of what it looks like. I always insert my plots this way and they always look exactly the same in the PDF as in MATLAB.
One bonus you also get with EPS is that you can actually specify a font size so that the text is readable even when you resize the image in the document.
As for the number of ticks, you can look at the axes properties in the MATLAB documentation. In particular, the XTick and YTick properties are very useful manually controlling how many ticks appear no matter what the window resolution is.
Edit (again): If you render the image to a raster format (such as PNG), it is preferable to choose the exact same resolution as the one used in the document. Rendering a large image (by using a big window size) and making it small in the PDF will yield bad results mainly because the size of the text will scale directly with the size of the image. Rendering a small image will obviously make for a very bad effect because of stretching.
That is why you should use a vector image format. However, the default MATLAB settings for figures produce some of the same problems as raster images: text size is not specified as a font size and the number of ticks varies with the window size.
To produce optimal plots in the final render, follow the given steps:
Set the figure's font size to a decent setting (e.g. 11pt)
Render the plot
Decide on number of ticks to get a good effect and set the ticks manually
Render the image to color EPS
In MATLAB code, this should look somewhat like the following:
function [] = nice_figure ( render )
%
% invisible figure, good for batch renders.
f = figure('Visible', 'Off');
% make plots look nice in output PDF.
set(f, ...
'DefaultAxesFontSize', 11, ...
'DefaultAxesLineWidth', 0.7, ...
'DefaultLineLineWidth', 0.8, ...
'DefaultPatchLineWidth', 0.7);
% actual plot to render.
a = axes('Parent', f);
% show whatever it is we need to show.
render(a);
% save file.
saveas(f, 'plot.eps', 'epsc');
% collect garbarge.
close(f);
end
Then, you can draw some fancy plot using:
function [] = some_line_plot ( a )
%
% render data.
x = -3 : 0.001 : +3;
y = expm1(x) - x - x.^2;
plot(a, x, y, 'g:');
title('f(x)=e^x-1-x-x^2');
xlabel('x');
ylabel('f(x)');
% force use of 'n' ticks.
n = 5;
xlimit = get(a, 'XLim');
ylimit = get(a, 'YLim');
xticks = linspace(xlimit(1), xlimit(2), n);
yticks = linspace(ylimit(1), ylimit(2), n);
set(a, 'XTick', xticks);
set(a, 'YTick', yticks);
end
And render the final output using:
nice_figure(#some_line_plot);
With such code, you don't need to worry about the window size at all. Notice that I haven't even showed the window for you to play with its size. Using this code, I always get beautiful output and small EPS and PDF file sizes (much smaller than when using PNG).
The only thing this solution does not address is adding more ticks when the plot is made larger in the latex code, but that can't be done anyways.
My code is simple enough, and importing eps images is something I've done before with other matlab-generated content, but for one reason or another I end up with blurred colors in my heatmap when I use epsfig or graphicx to import it into my document. In the picture below, the right is what shows up if I compile to DVI and open up the document in Yap, and the left is if I simply view the eps in GSView.
alt text http://img85.imageshack.us/img85/1694/epsproblem.png
Here is my code. This example is using graphicx, but the idea is the same with epsfig.
\begin{figure}
\centering
\includegraphics[scale=0.5]{images/ngram3_model_raw.eps}
\caption{The perplexity when compared
against the HUB test set}
\end{figure}
Is there perhaps some option I am forgetting?
I had the same problem with DVI, but if I compiled to pdf the pictures were fine
What you see is bilinear interpolation. It is done by the viewer. Probably Matlab defines the plot contents as a pixmap (I guess you use imagesc?).
The solution is not straightforward. It may help to use a different processing chain (as WtFudgE pointed out) that will lead the data end up in a format where it is not interpolated anymore. You may also use a different viewer that does not interpolate and I would assume that a printer would also not interpolate. This can again be dependent of the application you print from.
Sorry that I don't have a solution for you; at least you now have some new words to search for in Google. ;)