The Fingerprint Image Processing Based On Wavletlet Transform

In order to extract and match minutiae accurately in a fingerprint identification system, an effective enhancement method to destructed fingerprint images is extremely essential. With the development in fingerprint technology these years, a great number of fingerprint images reside in computers. So it is becoming a novel aspect to compress these images. Many scholars have done a lot of research on these subjects. However, there are no perfect solutions so far, due to the complexity of fingerprint applications. Another reason for the situation is that both computation complexity and precision have to be considered during the fingerprint processing. In this thesis, wavelet transform is applied to fingerprint image processing. Through wavelet transform, we enhanced and compressed fingerprint images in the frequency field, and achieve preferable results.After wavelet transform, an image is divided into several high frequency sub-images and a low frequency sub-image. The wavelet coefficients of these images indicate fine differences between different levels of resolution. There are several critical problems with wavelet transform: How to select an appropriate wavelet function and to choose the number of filters to perform the wavelet transform, and the levels of analysis are also to be considered. These problems were discussed in the enhancement research and compression strategy of fingerprint image respectively in the thesis.The distribution of wavelet coefficients has certain characteristics, we compressed and de-noised fingerprint images with these characteristics: Noises and the details of an image have dissimilar distribution characteristics under multi-resolution, so this characteristics can be used to design de-noising algorithm of image enhancement. Based on the characteristic of fingerprint ridge structure, we did further local directional compensation to the de-noised images, and got good results. As for image compression, based on the characteristic of wavelet coefficients, we applied special quantification strategy, which combined with the following Huffman coding, to complete the final compression.