Computed tomography is a useful tool for gas and plasma diagnostics. Unfortunately, viewing access in many tomographic experiments, such as in plasma physics, is strongly limited, which leads to a highly underdetermined inversion problem. A promising approach to this problem is the maximum entropy (ME) method. An application of the ME concept to the problem of two-view tomography is considered in this paper. It is shown that if additional prior information of the source is available it can be effectively incorporated into the maximum entropy algorithm. The crucial role of prior information is illustrated by several numerical examples.