Replacing scalar value with vector-value data

Hi,

I came across an image segmentation paper that provides a function similar to,

w=exp(x), where x is the intensity difference between 2 pixels.

It's mentioned that, x can be modified to |x| to handle vector-valued such as colored data.

I am not quite sure about the integration of the vector-valued data into the function.

If I am not wrong, for instance, a RGB image, the vector of pixel-1 is ([r;g;b])[234;12;31], and pixel-2 is [233;10;30]

http://imageshack.us/photo/my-images/163/er5y.jpg/

Therefore, to replace the x with |x|, does it simply mean, w=exp(xr+xg+xr) => w=exp(1+2+1) ?

Thanks in advance!

Answer 1

[Thorsten]

You can use the Euclidean distance between the two pixels

p1 = [234;12;31];
p2 = [233;10;30];
x = norm(p1 - p2);
w = exp(x)

[Lee ZY]

Thanks Thorsten!

Answer 2

[Lee ZY]

Thanks Thorsten! Is it a common practice to include vector-valued data?

and the concatenation method as in the previous post is not an acceptable way?

Thank you very much!