Python Image Processing Cookbook

How to do it...

Go through the following steps to implement the image color quantization with SOM:

1. Let's define the following function to segment an RGB color image with SOM. The function accepts the input image along with the dimension of the SOM grid (n_x x n_y) and the parameters σ and n (the number of random pixels to be chosen to train the network—that is, learn the weight vectors at the neurons). The image needs to be flattened (that is, each row should represent the RGB values for a single pixel) before the SOM is trained:

def segment_with_SOM(image, nx, ny, sigma=1., n=500): pixels = np.reshape(image, (image.shape[0]*image.shape[1], 3)) # SOM initialization and training som = MiniSom(x=nx, y=ny, input_len=3, sigma=sigma, \ learning_rate=0.2) ...