Numpy create an index of values in an array and replace values by others

Question

I have two sets of values RGB and XYZ in np.arrays that have the same length. The values in RGB and XYZ correspond to the same values in two different color spaces, the order of their appearance in the vectors is the same.

I have another array RGB_picture that contains values similar to the ones listed in RGB.

I want to find a way to replace the RGB values in RGB_picture by the corresponding XYZ values in a new array XYZ_picture.

example:

RGB = np.array([[0,0,0], [0,0,1], [0,1,0], [0,1,1], [1,0,0], [1,0,1], [1,1,1]])

XYZ = np.array([[0.237,0.32,0.8642], [0.43234,0.34,1], [0.2432,.31,0.324], [0.3242,1.324,1.234], [1.32,0.4,0.23], [1.432,0.32423,1.342], [1.234,1.324,1.324324]])

#6 pixels image
RGB_image = np.array([[[0,0,1],[0,0,1],[0,0,1],
                       [0,1,1],[1,0,0],[1,0,1]]])

"""I want to return this"""

>print XYZ_image 

>[[0.43234,0.34,1],[0.43234,0.34,1],[0.43234,0.34,1]],

>[0.3242,1.324,1.234], [1.32,0.4,0.23], [1.432,0.32423,1.342]]

I tried to do it with numpy but couldn't find how, would you help a bit?

Here's what I have but it is no numpy solution, so it is very slow:

XYZ_image = np.zeros_like(RGB_image)

w, h = RGB_image.shape

for i in xrange(w):
    for j in xrange(h):
        pixel = RGB_image[i,j]

        for k in xrange(len(RGB)):

            rgb = RGB[k]

            if rgb == pixel:

                XYZ_image[i,j] = XYZ[k]
                break

Can you show what you've tried so far, and explain specifically where you are stuck? — Cory Kramer
– Cory Kramer, Commented Jul 17, 2015 at 17:46

AQiccl135 · Accepted Answer · 2015-07-17 18:00:33Z

1

It sounds like you are trying to use the RGB values as indexes. You can solve this with regular Python. Make a dict of

my_dict = {RGB_value_1: XYZ_Value_1, RGB_value_2: XYZ_Value_2, etc}

then for every value in RGB_picture

my_dict[RGB_picture_value]

EDIT: Using your code

my_dict = {RGB[i]: XYZ[i] for i in range(len(RGB))}
return [my_dict[x] for x in RGB_image]

Please note that for this to work, RGB, and RGB_picture values need to be immutable (like tuples)

edited Jul 17, 2015 at 18:00

answered Jul 17, 2015 at 17:55

AQiccl135

331 silver badge7 bronze badges

Sign up to request clarification or add additional context in comments.

4 Comments

adrienlucca.net Over a year ago

immutable means it cannot be numpy arrays?

adrienlucca.net Over a year ago

I would really prefer a numpy solution, the arrays are big

AQiccl135 Over a year ago

Right now, your RGB values are lists RGB = np.array([[0,0,0], etc Lists can be changed, they are mutable. If you change them to tuples, RGB = np.array([(0,0,0), etc then they are immutable, and you can use them for indexes in a dict.

adrienlucca.net Over a year ago

yep, but the vector RGB_image is a numpy array

Collectives™ on Stack Overflow

Numpy create an index of values in an array and replace values by others

1 Answer 1

4 Comments

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

4 Comments

Related