压缩感知的指静脉图像去噪

Finger vein image denoising based on compressive sensing

基于压缩感知理论的梯度投影稀疏重建(GPSR)算法对合成指静脉图像进行去噪预处理,运用Canny算子提取指静脉边缘验证了GPSR算法的去噪效果。实验结果表明,与全变分去噪算法(ROF去噪算法)相比,运用GPSR算法可以得到更高信噪比的指静脉图像、更清晰的指静脉边缘轮廓,解决了红外传感器提取指静脉信息时存在的静脉边界模糊、不易分割及提取边缘等问题。

We present a Gradient Projection for Square Reconstruction （GPSR） algorithm for solving bound constrained quadratic programming problem to reduce the noise in synthetic vein images, which are blurred by various noises. This algorithm is based on the compressive sensingtheory. The edge of the vein was extracted by Canny operator to verify the GPSR method. Experiment results show that, compared with the total variation denoising algorithms developed by Rudin, Osher and Fatemi （also called ROF）, using the proposed GPSR algorithm can obtain finger vein image with higher Signal to Noise Ratio （SNR）, and clearer edge of the vein. So this algorithm can provide more accurate information for vein recognition and extraction.