Abstract: The value selection of expectation and variance in image normalization directly affects the quality of the image after normalizing and subsequent image processing. Through calculating a few of the fingerprint images’ gray boundary values in a set of images, this paper estimates the image expectation and variance, applies to the other fingerprint images. This can make the fingerprint image get a large dynamic range, not making the image gray scale limited, so as to realize the automatic estimation of image expectation. The experimental results show that this method is simple and reliable, and can keep the image gray levels, not reducing the contrast. This method is robust and can adapt to the different images, and it is a practical method of image normalization. Using this method to estimate image expectation automatically is beneficial to the improvement of automatic fingerprint identification algorithm and the related research of acquisition device independence.

Key words: image normalization, image enhancement, contrast, linear spanning

摘要： 图像规格化中的期望均值和方差值的选取，直接影响图像规格化后的质量和后续图像的处理。通过求取一组图像中少部分指纹图像的灰度边界值，对图像的期望均值和方差进行估计，并应用于该组中的其他指纹图像，可以使指纹图像获得较大的动态范围，而不至于使图像灰度受到限幅，从而实现对图像期望值的自动估计。实验结果表明，该方法简单可靠，能在不降低对比度情况下，保持图像的灰度层次。该方法鲁棒性好，能适应各种不同图像，是图像规格化的实用方法。采用该方法对图像的期望值进行自动估计，有利于自动指纹识别算法的改进以及采集设备无关性的相关研究。

关键词: 图像规格化, 图像增强, 对比度, 线性扩展