Abstract:Image mosaicking can transform a series of images to a large-view panorama, which has widely used in medical diagnosis and 3d reconstruction. Confronted with problems of low overlap and sparse texture, A quick mosaicking algorithm for accurate results was proposed. On the one hand, a new phase correlation method was designed based on hierarchical model, which dealt with low overlap and sparse texture well. Further, this algorithm was robust to illumination variation. On the other hand, a global optimization for accurate registration was proposed. This method eliminate unreliable offsets and avoid ghost phenomenon. Experimental results prove the algorithm has high accuracy and speed, which is suitable of image mosaicking with low overlap and sparse texture.