Flattening of numpy array

Question

I'm performing data mining on images. Each pixel is treated as a data point. The image is read as follows:

im=Image.open('lena.bmp')
im=numpy.array(im)
print im.shape

Depending on whether the image is color or grayscale, im.shape is either (10,10, 3) or (10,10,1)

After that, the image is flattened to a feature matrix as follows:

if (10,10,3), then --->(100,3)

if (10,10,1), then --->(100,1)

How do I write a polymorphic function for this? My current approach is:

obs=reshape(im,(im.shape[0]*im.shape[1],1, im.size/(im.shape[0]*im.shape[1])))

that's true. I'm having obs=reshape(im,(im.shape[0]*im.shape[1], im.size/(im.shape[0]*im.shape[1]))) — Dzung Nguyen
– Dzung Nguyen, Commented Feb 13, 2013 at 23:49

Jaime · Accepted Answer · 2013-02-13 21:51:55Z

3

You can do:

obs = np.reshape(im, (-1, im.shape[-1]))

answered Feb 13, 2013 at 21:51

Jaime

67.7k19 gold badges128 silver badges164 bronze badges

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

2 Comments

Can you please elaborate this code? It works perfectly. shape[-1] is the number of dimensions, how about -1 before it?

@DzungNguyen -1 is replaced internally by numpy to whatever value is needed to fit the size of the array.