关键词: 小波包变换, 指纹图像压缩, 无损差分脉冲编码调制（DPCM）, 嵌入式零数编码（EZW）, 分级压缩

Abstract: Aiming at the fingerprint image characteristic of plentiful medium frequency, and oppositely less high frequency and low frequency, the fingerprint image level compression algorithm is presented based on wavelet packet transformation. The fingerprint image is divided into three levels according to energy distribution after wavelet packet transformation. The medium frequency image, which contained more energy, is encoded by lossless Differential Pulse Code Modulation（DPCM）. The high frequency and low frequency images, which contained less energy, are encoded by Embedded Zero-tree Wavelet（EZW）. The image reconstruction combines the compressed image code streams with the feature point information. The experimental results indicate that compared with the traditional Zero-tree Wavelet encoding, the presented algorithm improves the average compression ratio about 1.832 and the average signal-to-noise about 4.07, meanwhile, reduces the average computation time by about 26% on the basis of ensuring the reconstruction quality, which proves the validity of level compression algorithm.

Key words: wavelet packet transformation, fingerprint image compression, Differential Pulse Code Modulation（DPCM）, Embedded Zero-tree Wavelet（EZW）, level compression