A direct method for optimization of a class of linear distributed parameter systems with stochastic distributed disturbances is presented. The method is based on expanding control systems, Green's function and covariance matrices in terms of a finite number of elements taken from complete sets of orthonormal basis in both time and space. Practical considerations are considered in the application of the method. A minimum-variance filter is designed using noisy discrete observations from a finite number of measuring points. Then the synthesis of a pointwise controller is carried out to minimize a quadratic performance index. The optimal coefficients of control are determined via simple algebraic relations in terms of the state conditional mean. A discussion of the properties and computational aspects of the method is given.