Fingerprint orientation field describes the essential texture features in a fingerprint impression, including shape, structure and direction. A representation scheme is introduced to describe the local orientation field around a minutia, so as to describe and discriminate different minutiae based on the neighbor texture information. On poor-quality fingerprint images, especially when heavy nonlinear distortion exists, minutia direction is no longer reliable enough. As a result, it is not effective to align two local orientation field images with a minutiae pair as the reference. An improved alignment and matching method of local orientation field images is proposed, which can achieve much better performance and improve the matching accuracy more clearly. Experimental results show that the proposed method can evaluate the resemblance between two minutiae more effectively.