Abstract: A novel algorithm for incomplete fingerprint recognition is proposed in this paper using fusion features and pattern entropy based similarity measure. Because of incomplete fingerprints unique characteristic of information loss, the recognition performance is mainly restricted by two critical problems: extracting features containing sufficient information and measuring similarity more effectively. For the first problem, minutiae and orientation field features are fused to get more comprehensive information and to improve the scale and rotation invariability. For the second, the pattern entropy is introduced to measure the coherency of correspondences between two feature sets to eliminate false match. The extensive experiments are done and compared with existing method on fingerprint databases and made thorough comparisons. Experimental results show that the proposed scheme has more efficient ability on separating genuine and impostor pairs and performs well in both accuracy and speed.