First, changing channels is very simple you just need to change the third dimension of the image, e.g.
Now, let's change channels from RGB to BGR:
image2(:,:,1) = image1(:,:,3);
image2(:,:,2) = image1(:,:,2);
image2(:,:,3) = image1(:,:,1);
Now in terms of the filtering with a Gaussian, if you apply a 3x3 to every channel, that would be the same if you do in RGB or in BGR as you are averaging pixels in a channel.
Hope that answers your question.