RGB image format doesn't follow the standard numpy RGB format

[GoogleCodeExporter]

write_image() supports RGB image if write_rgb=True is passed, but it expects an 
numpy array in the format 3xMxN for an image of MxN pixels, while the normal 
RGB images in numpy are encoded MxNx3. 

For reference, see the documentation here:
http://docs.scipy.org/doc/scipy/reference/generated/scipy.ndimage.imread.html

As a workaround, I currently send a view of my array with axes swapped to 
write_image. Like this:
image = scipy.ndimage.imread("test.jpg")
image_sw = image.swapaxes(2,0).swapaxes(2,1)
tif.write_image(image_sw, write_rgb=True)


I can try to provide a patch if you're interested.

Original issue reported on code.google.com by eric.p...@gmail.com on 9 Oct 2012 at 9:03

[GoogleCodeExporter]

Please find here a patch (against the current svn HEAD) to fix this issue. It 
enhances the behaviour of TIFF.write_image() when write_rgb=True. 

Basically, when this flag is True, write_image() will always write a single 
image, and try its best to describe it as a RGB (or RGBA) image.

Note that it also modifies the orientation tag to the correct value, which 
fixes issue 18.

The patch is available also as a git commit here:
https://github.com/delmic/pylibtiff/commit/cc38e51ae0857b605c81e7f631279b5cd54de
55d

Original comment by eric.p...@gmail.com on 5 Aug 2013 at 2:08

Attachments:

• 0002-improve-support-for-RGB-writing.patch

[GoogleCodeExporter]

Thanks! Patch applied to svn.

Original comment by pearu.peterson on 13 Sep 2014 at 10:11

• Changed state: Fixed