This paper presents a face recognition algorithm that addresses two major challenges. The first is when an individual intentionally alters the appearance and features using disguises, and the second is when limited gallery images are available for recognition. The algorithm uses a dynamic neural network architecture to extract the phase features of the face texture using 2D log polar Gabor transform. The phase features are divided into frames which are matched using the Hamming distance. The performance of the proposed algorithm is evaluated using three databases that comprise of real and synthetic face images with different disguise artifacts. The performance of the algorithm is evaluated for decreasing number of gallery images and various types of disguises. In all cases the proposed algorithm shows a better performance compared to other existing algorithms. © 2007 Elsevier B.V. All rights reserved.