In this paper, we propose a construction of an affine transformer for rectangular video image based on two transform sequences that are equivalent but composed of different steps. It comprises two identical blocks connected in parallel, each of which can transform each frame in two frame periods and consists of two types of modules. As an example, we show constructions of them. Especially, we show a simple configuration of address generators which need to generate complex address sequences for switching two transform sequences. The proposed affine transformer reduces hardware volume without reducing its speed compared with the existing one.