if you have stats toolbox, there is a function called kurtosis. Even otherwise, this should be very straightforward.
function k = kurtosis3(x,dim)
% Kurtosis function
mu = mean(x,dim);
repsiz = ones(1,length(size(mu)));
repsiz(dim) = size(x,dim);
d = (x-repmat(mu,repsiz)).^2;
n = d.^2;
k = mean(n,dim)./(mean(d,dim)).^2;
