Effects Of WSQ Compression On Client-Server Biometric Identification Performance

摘要

Indivos Corporation is building a large-scale client-server biometric authentication service to be utilized at the point-of-sale by leading retail merchants. The authentication service requires an enrollment step where five fingerprint images must be compressed and transferred to a centralized data processing center over existing network lines. The centralized data processing center will extract the minutiae from the uncompressed images. The obvious compression algorithm of choice is Wavelet Scalar Quantization (WSQ). The point-of-sale authentication step does not require WSQ compression, as a thin-client will process and extract the minutia points from the raw images. Algorithm performance data using this client-server model at various levels of WSQ compression are presented showing a degradation of match performance with increasing levels of compression.