Abstract: A seamless IR image sequence mosaic method based on feature points and principle of overlap transition Poisson image fusion was proposed. Firstly, a simplified SIFT was used to extract invariant feature points from images. In order to improve the accuracy of matching, bidirectional and complementary matching method was used. Then, random sample consensus(RANSAC) was used to perform reliable matching parameters of the transformation between images which were obtained from the matched feature points. Finally, Poisson image fusion was used to accomplish seamless image mosaic. The feature points were invariant to affine transformation, noise contamination, leading to robustness of the method. Experimental results show that the image mosaic method is simple and effective, and eliminate the mosaic seams while keeping good image definition.