How to process 3d gaussian convolution with a 3d array or matrix in OpenCV

I have a 3d array like this.

int sz[] = {240, 240, 240};
Mat accumarray(3, sz, CV_32S, Scalar::all(0)) ;

And I want to process a 3d gaussian convolution on this array to blur the array. To make data in the array somehow be cont开发者_运维知识库inuous.

However OpenCV only has 1d or 2d dft(), filter2d(), convolve().

Then how to process a 3d convolution in OpenCV? and how to build a 3d gaussian kernel efficiently? and how to separate the 3d convolution into multiple 2d or 1d convolution?