How to save images?

[aurelienbl]

Hello,

Thank you very much for this great package.
I have a very simple question. Probably to simple that I could not find the solution in the documentation.

How can I save image on disk in a proper format (png, jpg...).
More precisely, I have followed the colour-correction tutorial, and it is not clear to me how you save the corrected image in the end.

A bit of help would be appreciated!

Best,
Aurélien

[KelSolaar]

Hi @aurelienbl,

Simply passing the filename=my_image_name.png kwarg to most of the plotting definitions should be enough!

Cheers,

Thomas

[aurelienbl]

Hello Thomas,

Thank you for your replied, but I still have inconsistent result.

After the correction is done, I do:
colour.io.write_image(corrected_image, 'corrected_image.png')

Here is the result I have:

On the left, the corrected as displayed by the tutorial within the python interpreter. On the right the saved image.
I don't doubt that I have something wrong, but I don't see what!

Cheers,
Aurélien

[KelSolaar]

Oh sorry, I did not read carefully and you are not using the plotting code.

In order to write a PNG or JPG, i.e. 8-bit data, image to disk, you would need to non-linearly encode the image first, e.g. my_image = colour.cctf_encoding(my_image) then you can write it.

Is it possible to let me know which image IO backend you are using by giving me the output of the following definition: colour.utilities.describe_environment()

Cheers,

Thomas

[aurelienbl]

Hum I think I tried that, Here is corrected png.

(for reference I did
colour.io.write_image(colour.cctf_encoding(corrected) ,'corrected.png'))

Here is the output of colour.utilities.describe_environment():
defaultdict(collections.OrderedDict, {'Interpreter': OrderedDict([('python', '3.6.5 (default, Apr 25 2018, 14:23:58) \n[GCC 4.2.1 Compatible Apple LLVM 9.1.0 (clang-902.0.39.1)]')]), 'colour-science.org': OrderedDict([('colour', '0.3.15'), ('colour-checker-detection', 'v0.1.0-56-g246e441')]), 'Runtime': OrderedDict([('imageio', '2.5.0'), ('matplotlib', '3.1.1'), ('networkx', '2.2'), ('numpy', '1.16.4'), ('pandas', '0.25.3'), ('scipy', '1.1.0'), ('six', '1.11.0'), ('opencv', '4.2.0')])})

Thanks for the help!

[KelSolaar]

Looks like the image is compressed somehow! Let me get back to you tonight.

Bests,

Thomas

[KelSolaar]

Allright, took 5 minutes to look at that, it seems like imageio (PIL?) is trying to be smart and whenever it has floating point data casted to 8-bit, it remaps it as follows: [min, max] --> [0, 255] thus you should do something like that:

my_image = (colour.cctf_encoding(np.clip(my_image, 0, 1)) * 255).astype(np.uint8)

[aurelienbl]

Great ! It works fine like this.
Thank tou so much for the help.

Aurélien

[KelSolaar]

You are welcome! :)