Dear JImmy Gao,

Your question is a very important and easy to solve. normally SOM maps n dimensions data into 2 or 1 dimensions. 

convert each image to vector  e.g. v1 = reshape(im(:,:,1),ro*cl,1); ro is row size and col column size 

from v1 to vn assuming you have n images  vn = reshape(im(:,:,n),ro*col,1)

then use net = newsom(double([v1' v2' v3' . ..   vn']),[L P]); where L*P is the size of SOM network

net.trainParam.epochs = 20; % number of iterations

[net,trn,Y] = train(net,double([v1' v2' v3' . .. vn']));

Then Y is the data which contain your patterns

Good luck :)

Hi 

M.Awad

Thank you very much.I will have a try.

Regards

Jimmy

Hi 

M.Awad

When I established my self organzing map with images,the following mistake has happened.

My each image data has some missing data instead of NaN.

Is this have some problems to creat a som.I am very confused.

Here are four images to test som.

Hi. M.Awad. Could you please tell me how to display Y as an image in matlab?

Thank you very much.