[Numpy-discussion] Creating a RGB-image from BW

Zachary Pincus zachary.pincus@yale....
Tue Apr 28 06:47:38 CDT 2009


Hi Johannes,

According to http://www.pygtk.org/pygtk2reference/class- 
gdkpixbuf.html , the pixels_array is a numeric python array (a  
predecessor to numpy). The upshot is that perhaps the nice  
broadcasting machinery will work fine:

pb_pixels[...] = fits_pixels[..., numpy.newaxis]

This might not work, though. But perhaps it would be possible to make  
a numpy array that's really just a view onto the memory of pb_pixels,  
or perhaps one could convert fits_pixels into a numeric array...  
Hopefully someone on the list can make a suggestion about dealing with  
numeric arrays.

Alternately, there are pixbuf methods for reading image data from  
strings. You'd just need to get fits_pixels set up properly, then call  
tostring() on it, and pass that to the pixbuf. The trick is in setting  
up fits_pixels so that its memory layout corresponds to what gtk  
wants. Usually, images are stored in memory as (r,g,b) tuples packed  
by rows and then columns; this is I assume what GTK wants. So you'd do  
something like:

fits_color = numpy.empty((height, width, 3), dtype=numpy.uint8)
fits_color[...] = fits_pixels[..., numpy.newaxis]
fits_string = fits_color.tostring()
pb = gtk.gdk.pixbuf_new_from_data(fits_string, gtk.gdk.COLORSPACE_RGB,  
False, 8, 640, 480, 3*640)

Zach

On Apr 28, 2009, at 2:36 AM, Johannes Bauer wrote:

> Hello group,
>
> I've been redicted from usenet ("Convert numpy.ndarray into "normal"
> array", <75dgm1F16hqntU1@mid.dfncis.de>) here and hope this is the  
> right
> place.
>
> Basically, what I have is a numpy-Array which I got from a FITS-file
> (it's black/white). I want to display that using GTK. Therefore every
> color needs to appear three times (to make it look gray R = G = B).
>
> The basic framework looks like
>
> [...]
> pb = gtk.gdk.Pixbuf(gtk.gdk.COLORSPACE_RGB, False, 8, width, height)
> pb_pixels = pb.get_pixels_array()
>
>
> print(type(pb_pixels), pb_pixels.shape, pb_pixels.typecode())
> print(type(fits_pixels), fits_pixels.shape, fits_pixels.dtype)
>
> which gives
>
> (<type 'array'>, (480, 640, 3), 'b')
> (<type 'numpy.ndarray'>, (480, 640), dtype('uint8'))
>
> so now I need to assign values. Naively I started out with
>
> for x in range(width):
> 	for y in range(height):
> 		pb_pixels[y, x] = fits_pixels[y, x]
>
> which was horribly slow (around 3 seconds). Thanks to the usenet  
> help, I
> now got somewhat better:
>
> fits_colors = numpy.zeros((height, width, 3), dtype="uint8")
> for y in range(height):
> 	for x in range(width):
> 		fits_colors[height -  y - 1, x] = fits_pixels[y, x]
> pb_pixels[:, :] = fits_colors
>
> This also works, and is a lot faster (around 0.7 seconds). However,
> there seems to be a better way to do it. I played around with
>
> fits_colors = numpy.fromfunction(lambda y, x, z: fits_pixels[y, x],
> (height, width, 3), dtype="uint8")
> pb_pixels[:, :] = fits_colors
>
> Which worked somewhat - but gives weird results: The picture is
> rotatated 90° to the right and the lower left part is displayed
> repeatedly after 256 pixels... (I can make a screenshot if that's
> easier). The fromfunction Function is quite fast in my context (around
> 0.2 second).
>
> How should I solve this problem the right way?
>
> Kind regards,
> Johannes
> _______________________________________________
> Numpy-discussion mailing list
> Numpy-discussion@scipy.org
> http://mail.scipy.org/mailman/listinfo/numpy-discussion



More information about the Numpy-discussion mailing list