How to extract a color channel from array of images in python

Question

I have a tensor of 10 samples, each which contain 10 time-series 20x20x3 RGB images which I would like to extract the green color channel

The images are stored in an array of arrays called images

For example:

images[0][0][:,:,1]

returns back the green channel for for one image in one sample.

However, when I try to use the command:

images[0][:][:,:,1] 

I receive the error:

IndexError: too many indices for array

How would I generalize my first line of code to pull all of the green channel images from the 1st sample?

Shapes of the data:

images.shape
(10,)

images[0].shape
(10,)

images[0][0].shape
(20,20,3)

Here is a sample of the data. The data are images which were extrated from a .mat file so it is stored as an array of arrays, with a sample shown below:

images
array([[array([[[41,  0,  0],
    [43,  0,  0],
    [45,  0,  0],
    ...,
    [18,  0,  0],
    [ 5,  0,  0],
    [ 0,  0,  0]],

   [[45,  0,  0],
    [50,  0,  0],
    [49,  0,  0],
    ...,
    [ 3,  0,  0],
    [ 0,  0,  0],
    [ 0,  0,  0]],

   [[49,  0,  0],
    [49,  0,  0],
    [48,  0,  0],
    ...,
    [ 0,  0,  0],
    [ 0,  0,  0],
    [ 0,  0,  0]],

   ...,

   [[16,  0,  0],
    [ 5,  0,  0],
    [ 1,  0,  0],
    ...,
    [ 0,  0,  0],
    [ 0,  0,  0],
    [ 0,  0,  0]],

   [[ 3,  0,  0],
    [ 1,  0,  0],
    [ 0,  0,  0],
    ...,
    [ 0,  0,  0],
    [ 0,  0,  0],
    [ 0,  0,  0]],

   [[ 0,  0,  0],
    [ 0,  0,  0],
    [ 0,  0,  0],
    ...,
    [ 0,  0,  0],
    [ 0,  0,  0],
    [ 0,  0,  0]]], dtype=uint8),
    array([[[87,  0,  0],
    [92,  0,  0],
    [86,  0,  0],
    ...,
    [33,  0,  0],
    [51,  0,  0],
    [60,  0,  0]],

   [[90,  0,  0],
    [88,  0,  0],
    [79,  0,  0],
    ...,
    [11,  0,  0],
    [21,  0,  0],
    [41,  0,  0]],

   [[89,  0,  0],
    [82,  0,  0],
    [62,  0,  0],
    ...,
    [12,  0,  0],
    [ 4,  0,  0],
    [16,  0,  0]],

   ...,

   [[77,  0,  0],
    [77,  0,  0],
    [76,  0,  0],
    ...,
    [48,  0,  0],
    [44,  0,  0],
    [42,  0,  0]],

   [[88,  0,  0],
    [85,  0,  0],
    [85,  0,  0],
    ...,
    [54,  0,  0],
    [53,  0,  0],
    [51,  0,  0]],

   [[89,  0,  0],
    [89,  0,  0],
    [88,  0,  0],
    ...,
    [55,  0,  0],
    [54,  0,  0],
    [53,  0,  0]]], dtype=uint8),

Answer 1

Something like this?

green_0 = [i[:,:,1] for i in images[0]]
all_greens = [[i[:,:,1] for i in ims] for ims in images]

but if images is already a numpy array, you can simply do images[0,:,:,:,1] and images[:,:,:,:,1]

Answer 2

Assuming that images is a NumPy ndarray with the following dimensions:

images[sample_dim, time_dim, width, height, color]

you could simply resort to a single slicing operation, e.g.:

images[:, :, :, :, 1]

to get only green across your dataset.

---

What you have been doing, i.e.:

images[0][0][:, :, 1]

could be rewritten more cleanly and efficiently as:

images[0, 0, :, :, 1]

---

To understand why images[0][0][:, :, 1] does work and images[0][:][:, :, 1] does not, I suggest you to take a look at the shapes of images[0][0] and images[0][:], which are the object you are trying to slice the final time with [:, :, 1]

---

If images is not a single multidimensional array, you could just call np.array() constructor on nested lists and it should convert it to a proper ndarray.

---

Instead, if your input is a NumPy array of objects, where each object is also a NumPy array, the most efficient approach is:

np.array(images.tolist())