In this paper, we propose a new scheme that a user enrolls his fingerprint images sequentially captured by rolling and sliding his finger, thus continuously contacting on the sensor. We also developed an image-fusion algorithm to mosaic the images obtained by the enrollment scheme. Conventional fusion algorithms for fingerprint images are based on large-sized sensors, and they are easily failed to combine images if there are not enough common areas among images. Our enrollment scheme assures that the common area between two sequential images is large enough to be combined even with a small-sized sensor. Experimental results show that average combined images are 1.91 times larger than a single image, and success rate for combining is 2.3 times higher than a conventional dab approach.