In this paper, we present an optimization mechanism for two popular landmark-based mobile robot visual homing algorithms (ALV and HiSS), called vector pre-assigned mechanism (VPM). VPM contains two branches, both of which can promote the homing performance effectively. In addition, to make the landmark distribution satisfy the equal distance assumption, a landmark optimization strategy is proposed based on imaging principle of the panoramic vision. Experiments on both panoramic image database and a real mobile robot have confirmed the effectiveness of the proposed methods.