基于区域能量加权的图像融合算法在指纹图像提取中的应用

Application of Image Fusion Algorithm Based on Regional Energy Weights in Fingerprint Image Extraction

在拍摄光滑客体表面的无色指纹时,一般采用定向反射光照相的方法,但是对于光滑纸板等表面,潜在指纹与背景之间的反射亮度差很小,传统方法难以进行拍摄提取。使用不同方向的光源配光,每一方向光源采集一幅图像,然后使用小波变换,将图像分解为不同尺度空间上的一系列子图像。对于低频子图像,采用取平均的融合规则;对于高频子图像,采用基于窗口的区域能量决策进行系数融合。最后对融合后的图像进行基于自适应引导滤波的多尺度Retinex图像增强,以增加指纹图像对比度。实验结果证明,该算法可以解决光滑物面指纹图像难以提取的问题,从而得到较为清晰的指纹图像。

When taking a colorless fingerprint on the surface of a smooth object, we usually adopt the method of directional reflec ted light photography. However, for glossy cardboard, the difference in reflection brightness between the potential fingerprint and the background is small, and the traditional method is difficult to extract. We use different directions of the light distribu tion, to capture an image at each direction of the light source. Then by using wavelet transform, the image is decomposed into a series of sub images in different scales. For low frequency sub images, the average fusion rule is adopted. For the high fre quency sub images, window based regional energy weighting is used to perform coefficient fusion. Finally, the multi scale Ret inex image enhancement with adaptive guided filtering of the fused image is used to increase the contrast of the fingerprint im age. Experimental results show that this algorithm can solve the problem of that it is difficult to extract the smooth object sur face fingerprint image so that a clear fingerprint image can be ottained.