Multi-scale Palmprint ATD Points Extraction And Adaptive Direction Field Smoothing

Biological studies have shown that the ATD angle of the palmprint has a great correlation with the possibility of female breast cancer.However,the existing algorithms for extracting palmprint ATD angles are relatively few and many of them are limited to artificial markers and manual measurements,which has the drawback of low accuracy and repeatability.Therefore,the purpose of this paper is to accurately locate the positions of the palmprint a,t,and d points and calculate the size of the ATD angle.The main contribution of the paper is as follows:(1)A method of slicing the ATD position instead of directly processing the palmprint is proposed.Firstly,the image is normalized.Subsequently the modified disk method is used to locate the root point of the finger on the accurate palm contour.Finally,the positional characteristics of the ATD point of the palmprint are analyzed and the position area is reduced.The experimental results show that the proposed method can reduce the area of the ATD points and the noise contained in the position slice is less,which greatly simplifies the complexity of subsequent palmprint processing.(2)A method to improve the palm line contrast of valleys and ridges based on Retinex theory is proposed.Firstly,based on the Retinex theory,the ray components of the human eye recognition object are analyzed and;then the computational complexity is simplified by the logarithmic domain transform,and the Gaussian low-pass filter is designed to remove low-frequency components from the image,which is useful for the image contrast.High-frequency components are remained;Finally,the enhanced valley and ridge contrast is obtained by inverse transform and stretching the pixel grayscale values.The experimental results show that the proposed method can not only improve the ridge line contrast of non-contact palmprint images,but also balance the influence of light.(3)An algorithm for building a palmprint direction field based on a composite block is proposed.Firstly,the directional based on gradient method is introduced;Subseqently the noise-contaminated area is judged by adaptive consistency,and the direction of the noise area is recalculated by the correct direction of its neighborhood,and the noise area is reduced through continuous iteration to achieve smoothness.Finally,the direction field obtained by the different direction modes of different Gaussian kernel scales is smoothed.The experimental results show that the proposed algorithm not only has good effect on the smoothness of noise-polluted areas,but also has a good effect on the removal of.(4)A multi-scale ATD point detection algorithm for complex filters is proposed.Firstly,the directional field under different scales of composite block is calculated and smoothed by filtering.Subsequently,an improved Pointcare Index(PI)value detection algorithm is used to detect the directional field of different scales from the large scale to the small scale.And the detected triangle points are fitted;Finally a triangular symmetric complex filter is applied to average the neighborhood of all points to obtain the average of the amplitudes.The maximum average is corresponding to the final triangle position.The experimental results show that the proposed algorithm is accurate for detecting the ATD point of the palmprint slice,and basically does not produce misjudgment on the triangle point,and the triangle does not shift the genuine position.Finally,by statistically comparing the locations of the extracted triangle points by using a number of samples,experimental results prove that the accuracy of the triangle point extraction of the proposed method is 85%.For palmprint images,the location of the T point at the ATD is generally overlapped by main lines,which is indicated as noise of the image,and the quality of some images are greatly low.Therefore,human eyes are pretty difficult,but the proposed can effectively for T triangle point in low quality palmprint images.