基于局部特征的正则化滤波算法-计算机应用研究.pdfVIP

优先出版 计 算 机 应 用 研 究 第33 卷 基于局部特征的正则化滤波算法* 1, 2 1, 2* 1, 2 杨 昊 ，陈雷霆 ，邱 航 (1．电子科技大学 计算机科学与工程学院，成都 611731 ；2 ．数字媒体技术四川省重点实验室，成都 611731) 摘 要：大部针对脉冲噪声的去噪算法主要利用预设的全局阀值来对含噪图像进行恢复，因此对图像局部边缘和细节的 恢复能力较差。对此，提出一种基于图像边缘特征的正则化滤波算法。算法首先通过反映图像边缘细节的局部统计量建 立自适应阀值，用以进行脉冲噪声检测，得到噪声候选集合；然后以此为基础，优化保边正则化方法，提升算法性能。 仿真实验从定量数值比较和主观视觉比较两个方面表明，本算法在噪声密度较高时，仍能提供较好的图像恢复效果，从 而为图像的后续研究提供了良好的基础。 关键词：脉冲噪声；自适应阀值；保边正则化；滤波 中图分类号：TP391.41 Regularization filtering algorithm based on local characteristics Yang Hao1, 2, Chen Leiting1, 2*, Qiu Hang1, 2 (1. College of Computer Science Engineering, University of Electronic Science Technology of China, Chengdu 611731, China; 2. Sichuan Provincial Key Laboratory of Digital Media, Chengdu 611731, China) Abstract: Most algorithms for removing impulse noise adopt the predefined global threshold to restore the corrupted images, therefore, they are not good at the recovery of local edges and details in images. Thus, this paper provides a regularization filtering algorithm based on edges’ characteristics. The algorithm builds an adaptive threshold by the local statistics which can reflect the edges and details in images. These thresholds are adopted to detect noisy pixels and get a candidate set filled of noise. Then the candidate set is used to optimize the edge-preserving regularization method for its performance. Simulation results show that proposed algorithm can still provide better performance of restoration by quantitative comparison and subjective visual comparison when noise level is high, thereby the algorit