Half-orientation extraction of palmprint features

• The points in palmprint images usually have two different dominant orientations.
• Half-orientations can precisely represent the orientation feature of the palmprint.
• Defining a bank of “half-Gabor” filters for the half-orientation extraction.
• Our method improves the performance of palmprint recognition.

Orientation features of the palmprint are usually used in palmprint recognition methods. Conventional orientation based methods are always based on an assumption that the line in a palmprint is straight and possesses only a single dominant orientation. However, a large number of “lines” in a palmprint are curves. The point in these curves usually has two dominant orientations. Moreover, it can be seen that there are numerous cross wrinkles in a palmprint. The cross point of any two cross wrinkles obviously has two different dominant orientations. In this paper, we proposed a simple and effective double half-orientation based method for feature extraction and recognition of the palmprint. In the method, a bank of “half-Gabor” filters are defined for the half-orientation extraction of a palmprint. Compared with the single dominant orientation, the double half-orientations can more precisely characterize the global orientation feature of a palmprint. Extensive experiments are carried out on three different kinds of palmprint databases and the results show that the proposed method achieves a promising performance in both palmprint verification and identification and outperforms other orientation feature based methods.