This paper presents a unified framework for control structure selection of multivariable processes. On the basis of structure decomposition, all interaction transmitting channels to an individual control loop from all the other closed loops are decoupled and represented by elements of the decomposed relative interaction array (DRIA). Utilizing the decoupled interactions from both inputs and outputs, a decomposed relative gain array (DRGA) is constructed. DRGA is then employed to evaluate the overall interactions to individual control loop from another one. Consequently, the synthesis of the control structure can be evaluated in a straightforward manner by using a relative control performance index (RCPI). A simple Cai-He (C-H) diagram is proposed to facilitate control structure selection procedure. Two examples are presented to illustrate the advantages of the proposed approach.