This paper presents a method to combine so|called multiband DFT filter bank with bivariate Box splines to construct a perfect reconstruction model for image processing. Based on the theory of B|splines

which are useful tool in many fields

Box Spline is said to be an extension

with the aim at describing multi|variate functions. In this paper

firstly

it introduces the DFT filter bank

and presents the condition to form a perfect reconstruct model. Then it describes bivariate Box splines and its some properties in little more details. It especially presents a particular class of bivariate Box splines

and use it to construct a decomposition/reconstruction model

which may be candidates for some kinds of task in image processing. This model is proved to be satisfied by the perfect reconstruction condition

and thus can be quite useful theoretically. At last

some experimental results are given to demonstrate the use fulness of this model.