6

How do I convert an M x N grayscale image, or in other words a matrix or 2-D array, into an RGB heatmap, or in other words an M x N x 3 array?

Example:

 [[0.9, 0.3], [0.2, 0.1]]

should become 

[[red, green-blue], [green-blue, blue]]

where red is [1, 0, 0], blue is [0, 0, 1],etc.

rd11
  • 3,026
  • 3
  • 23
  • 32
  • Also see: http://stackoverflow.com/questions/14869321/is-there-a-way-to-convert-pyplot-imshow-object-to-numpy-array/14877059#14877059 . Remember to accept your own answer when it will let you. – tacaswell Mar 15 '13 at 03:32
  • I agree, this is similar to that question, although this is worded more clearly. (I searched for a while and couldn't find that question.) I'm pretty new to stackoverflow; do we merge or something? – rd11 Mar 15 '13 at 12:41
  • Don't worry about it, you need to take no action. I have flagged it as a possible duplicate, if 4 other people with 3k+ rep agree this will get closed as a duplicate (which just means no new answers, and a permanent link to the other question). If other people don't agree, the comment will stay, but my close vote will go away. I agree you did a better job of identifying the real problem (based on the comment in your answer about using `imshow`). – tacaswell Mar 15 '13 at 14:44

1 Answers1

19
import matplotlib.pyplot as plt

img = [[0.9, 0.3], [0.2, 0.1]]

cmap = plt.get_cmap('jet')

rgba_img = cmap(img)
rgb_img = np.delete(rgba_img, 3, 2)

cmap is an instance of matplotlib's LinearSegmentedColormap class, which is derived from the Colormap class. It works because of the __call__ function defined in Colormap. Here is the docstring from matplotlib's git repo for reference, since it's not described in the API.

def __call__(self, X, alpha=None, bytes=False):
    """
    *X* is either a scalar or an array (of any dimension).
    If scalar, a tuple of rgba values is returned, otherwise
    an array with the new shape = oldshape+(4,). If the X-values
    are integers, then they are used as indices into the array.
    If they are floating point, then they must be in the
    interval (0.0, 1.0).
    Alpha must be a scalar between 0 and 1, or None.
    If bytes is False, the rgba values will be floats on a
    0-1 scale; if True, they will be uint8, 0-255.
    """

A simpler option is to display img, using plt.imshow or plt.matshow, and then copy or save the result as an RGB or RGBA image. This was too slow for my application (~ 30 times slower on my machine).

rd11
  • 3,026
  • 3
  • 23
  • 32
  • +1. With the described method (instead of imshow) you are guaranteed to get back an image with same dimensions. If you want an image of different size, then you can get it from imshow (with some tuning needed), since it already interpolates the image for you. – heltonbiker Mar 14 '13 at 20:45