Fingerprint Image Compression Based on Directional Filter Banks and TCQ

Abstract

A novel compression algorithm for fingerprint image is proposed. Wavelet-based Contourlet transform (WBCT), which is based on wavelet transform and directional filter banks (DFBs), can be capable of efficiently approximating natural images containing contours and oscillatory patterns. To reduce frequency scrambling, a new scheme based on maximally-flat filters which implements the DFBs is proposed. A quadtree sorting procedure, similar to SPIHT, is used to explicitly form classes of WBCT coefficients. The classes are encoded using arithmetic and trellis-coded quantization. The resulting encoding algorithm offers consistent improvement over SPIHT performance. Simulations show that the new encoding algorithm provides improved encoding performance over SPIHT and preserves more fingerprint image details.