For the purpose of solving pose estimation problem in robot hybrid visual servo applications, the fundamental properties of Euclidean homography matrix are analyzed, and a fast homography decomposition algorithm in the case of known normal to the scene plane is proposed. Experimental results show that the proposed algorithm can be simply implemented, and because of the avoidance of time-consuming singular value decomposition and the process of distinguishing the unique one from multiple solutions, the algorithm can well meet the real-time requirement of the visual servo system.