Fingerprint image compression algorithm based on matrix optimization

The fingerprint database calls for the valid compression algorithm of the storage and transmission. Wavelet Scale Quantization (WSQ) is commonly adopted for the general image compression, however, it dose not sufficiently take account of the ridge texture character. Above all, this paper proposes a matrix optimization model for the image compression and reconstruction (MOMICR). In succession, we put forward a two-dimensional ridgelet non-negative matrix factorization (TDR-NMF) algorithm to find the optimal solution of fingerprint MOMICR. Experiments both on FVC2002 database and hundreds of latent fingerprints randomly chosen from the databank supplied by Beijing Municipal Public Security Bureau show that the proposed method can get a good PSNR even without coding or other post-process, holding most minutiae.