Research Of Palmprint Recognition Based On Gabor Filters

Biological feature identification technology has uniqueness, stability and convenience, becomes the new darling of authentication technology, it is applied more and more widely. Palmprint identification has been focused on by many researchers due to its advantages such as easy capturing, abundant image information and high-rate acceptance by users etc. It becomes a very popular method of identification in recent years.But in the current palmprint recognition technology, how to effectively extract the region of interest, and represent palmprint features are the focus and difficulties in palmprint recognition. According to this, this paper combines the palmprint images'features and the Gabor filters'direction selection, do the following works:(1) At the palmprint image preprocessing stage, puts forward a new and simple extraction method based on searching palmprint's pixels. The traditional method has no universality, process is complicate, and the calculation is large. This paper's method is simple, small amount of calculation and effctivity.(2) Before the feature extraction, combine with the direction of palmprint, this paper presented two automatic methods for selecting the direction. One is based on the spectral characteristics of palmprint image, map it to polar coordinates domain, then the peak of spectrum on the projection axis is analyzed and calculated, at last the direction parameters of Gabor filter is obtained. The other is through the Radon transform to detect straight lines, from the transform domain to detect the line's direction and then set the direction of Gabor filters. Using empirical parameter to set direction ignored the characteristics of the image itself and the differences between classes; through the above two methods for filter's direction setting, used the palmprint texture and Gabor filter's direction selectivity well, makes the following part of extracting the palmprints'characteristic more effective.Based on the above work, this paper filter the palmprint image, extract mean and variance to be palmprint's features.In the matching stage, use the most simply NN (Nearest Neighbor) method to match the features. Specially, gives a detailed analysis and comparison between the traditional direction set method and parameters sets based on image content , showed the importance of filter direction parameters and the effectiveness of the methods proposed in this paper.