Motion estimation is one of the key technologies in inner-frame estimation encoding of video compression. In all kinds of compression standard, motion estimation based on block is used widely. The H.264 was a kind of new generation video compression standard. In order to reduce the computational complexity of the motion estimation and improve the accuracy of motion compensation, a new search algorithm based on campaign features is pointed out. An adaptive model for starting point prediction is presented. And an alterable search pattern is introduced to improve the searching efficiency by adjusting its shape according to the motion characteristics. Experimental results show the new algorithm has better image quality and speed of the search advantage.