Fingerprint Image Enhancement Processing Method Based on Weighted Median Filtering

Abstract

When the traditional median filtering method is used to filter the impulse noise of the fingerprint image, the filtering window size is always fixed. The traditional median filtering method does not take into account the distance correlation between the surrounding pixel points and the central pixel point in the filtering window. To some extent, the traditional median filtering method is less effective in filtering impulse noise, especially in filtering high-density impulse noise. In order to solve these problems of the traditional median filtering method, a method based on weighted median filtering is proposed for fingerprint image enhancement processing in this paper. In the method, the filtering window is first initialized. The center pixel of the filtering window is detected according to its pixel value to determine whether it is impulse noise. If the center pixel is not impulse noise, the current center pixel value is taken as the filtered output value. If the central pixel is impulse noise, the pixels in the filtering window are detected to determine whether all these pixels are impulse noise. If not all these pixels are impulse noise, the non-impulse noise pixel points in the window are regarded as valid pixel points, and the median value of the valid pixel points is taken as the filtered output value. If all the pixels in the window are impulse noise, the size of the filtering window is increased and it is detected whether the filtering window reaches the maximum value. At this time, whether all the pixels in the new filtering window are impulse noise is detected. If the filtering window size reaches the maximum value and all the pixels in the filtering window are impulse noise, the filtered pixel points in the filtering window are weighted to calculate the center pixel value of the filtering window. The proposed method is evaluated on fingerprint images contaminated by different density impulse noises, and is compared with the traditional median filtering method. The experimental results show the effectiveness and feasibility of the proposed method.