In fingerprint identification system, fingerprint image is influenced inevitably by noise. So minutia extracting and recognition rely heavily the result of preprocess which can decreases the noise. The normalization, which belong to preprocess operation, based on blocking process of fingerprint image which is proposed. First, an original image is divided into sub-blocks with the size W W and the region of interest of the fingerprint image is obtained. The desired mean and variance for the image normalization are determined according to property such as mean and variance of each block. The proposed algorithms are tested with FVC fingerprint database images and confirmed have better performance.