[Matplotlib-users] Exporting plot with given scale and DPI

Alexander Bruy alexander.bruy at gmail.com
Tue Jan 9 11:58:41 EST 2018


thanks for the hints, I have adopted your suggestions but still have
some issues.
Now my workflow looks like this:
1. get scale and DPI from user
2. calculate figure width and height in inches based on scale and DPI
3. make a figure with calculated width and height and small DPI value
    fig = plt.figure(figsize=(imgWidth, imgWidth), dpi=10)
4. plot profile line
5. export figure with correct DPI
    fig.savefig(fileName, dpi=int(userDPI), format="png", bbox_inches="tight")

Unfortunately, when I export figure with correct DPI I get incorrect image
dimensions. After some experiments, I have found that adding fig.show()
*before* the savefig() call fixes things and resulting image is correct.

Here is the image obtained without fig.show()
for example https://ibb.co/dx03E6

And here is correct result I get with fig.show() added before savefig()

Any ideas what can be wrong? I tried to remove bbox_inches="tight" but
still no luck, output is not correct anyway. Using mpl.use('Agg') or similar
things is not possible as matplotlib embeeded into Qt application.

I can live with fig.show() workaround, but need a way to close plot window
programmaticaly after saving the figure.

2018-01-08 21:29 GMT+02:00 Eric Firing <efiring at hawaii.edu>:
> You are correct that if you are displaying a plot on the screen, the figure
> size will be changed to fit.  This is a shortcoming of all the mpl backends;
> the plot windows should have automatic scrollbars instead. The workaround is
> simple, though: When you make the figure, use a small enough dpi so that the
> window will fit on the screen.  Then use your desired dpi when you save:
> fig, ax = plt.subplots(figsize=(20,30), dpi=20)
> # ... plot something...
> fig.savefig('bigplot.png', dpi=150)
> Assuming you have a typical screen size, the above will result in the
> following output from the "file" command (in Linux or OSX):
> bigplot.png: PNG image data, 3000 x 4500, 8-bit/color RGBA, non-interlaced
> As you see, it is consistent with the requested figsize and file dpi.
> Eric
> On 2018/01/08 7:45 AM, Alexander Bruy wrote:
>> It is not clear from the docs if this function preserves dimensions and
>> DPI
>> when exporting plot into PNG.
>> Also as I understand it resizes plot on screen and does not play well with
>> large sizes.
>> 2018-01-08 19:20 GMT+02:00 Jody Klymak <jklymak at uvic.ca>:
>>> Does `fig.set_size_inches()` do what you want?
>>> Cheers,   Jody
>>>> On 8 Jan 2018, at 08:44, Alexander Bruy <alexander.bruy at gmail.com>
>>>> wrote:
>>>> Hi all.
>>>> I use matplotlib inside PyQt app and want to export plot into PNG image
>>>> with
>>>> some user-defined scale and DPI. So after printing that image I should
>>>> be able
>>>> to measure on the paper according to the scale.
>>>> For example, length of the profile line is 500m (this is my x-axis),
>>>> scale set to 1:10000 and DPI is 150. 500m at 1:10000 should result in
>>>> a 5cm print out. Taking 150 DPI into account resulting bitmap should
>>>> be 295 pixels wide:
>>>> 1cm is 0.393701 inches
>>>> 5cm is 1.968505 inches
>>>> 1.968505inches * 150DPI = 295.27575 px
>>>> Also it is necessary to take into account that resulting image may be
>>>> really big, for
>>>> example if profile line length is several kilometers and DPI is 600.
>>>> As I can see there is a 'dpi' parameter in the savefig() call, but I
>>>> can't figure out how
>>>> to adjust size of the existing figure when exporting. Can you help me?
>>>> Thanks
>>>> --
>>>> Alexander Bruy
>>>> _______________________________________________
>>>> Matplotlib-users mailing list
>>>> Matplotlib-users at python.org
>>>> https://mail.python.org/mailman/listinfo/matplotlib-users
>>> --
>>> Jody Klymak
>>> http://web.uvic.ca/~jklymak/
> _______________________________________________
> Matplotlib-users mailing list
> Matplotlib-users at python.org
> https://mail.python.org/mailman/listinfo/matplotlib-users

Alexander Bruy

More information about the Matplotlib-users mailing list