CNOT gate application to separated qubits 
In the case 1 in the picture, it is easy to perform matrix calculations concerning the circuit to obtain a final state.
In case 2 however, I am wondering what is a general procedure to calculate it with matrices? None of tensor products of gates seems right. A decomposition of the CNOT gate into two matrices and sandwiching the Hadamard with them seems to be the only option but it is impossible to do it.
Can you help me?
 A: The basic approach is to break independent operations to different gates and then calculate the gate operation on all the basis vectors (it's not very pleasant but always work). At your case the representation matrix is - 
$ G=(\mathbb{I} \otimes \mathbb{H} \otimes \mathbb{I})(C_{1,3})    $  
$ C_{1,3}=\begin{pmatrix}
1 & 0 & 0 & 0 & 0 & 0 & 0 & 0 \\
0 & 1 & 0 & 0 & 0 & 0 & 0 & 0 \\
0 & 0 & 1 & 0 & 0 & 0 & 0 & 0 \\
0 & 0 & 0 & 1 & 0 & 0 & 0 & 0 \\
0 & 0 & 0 & 0 & 0 & 1 & 0 & 0 \\
0 & 0 & 0 & 0 & 1 & 0 & 0 & 0 \\
0 & 0 & 0 & 0 & 0 & 0 & 0 & 1 \\
0 & 0 & 0 & 0 & 0 & 0 & 1 & 0 \\
 \end{pmatrix} $
at basis -
$|\psi_1\psi_2\psi_3\rangle \rightarrow  \begin{pmatrix}
|000 \rangle \\
|001 \rangle \\
|010 \rangle \\
|011 \rangle \\
|100 \rangle \\
|101 \rangle \\
|110 \rangle \\
|111  \rangle \\
 \end{pmatrix} $
Notice the locations of the $ \mathbb{X} $ (NOT operator) in the $C_{1,3} $. Essentialy this is a $ \mathbb{CX} $ gate, on those qubits, and you can make it a $ \mathbb{CU} $ gate ( $ \mathbb{U} $ any unitary matrix ) by replacing $ \mathbb{X} $ with $ \mathbb{U} $ in the matrix.
