摘要
针对传统PM(Perona-Malik)滤波去噪模型在处理声纳图像时存在过度扩散和因参数设置不当导致方程畸变等问题,提出了一种融合多尺度自回归MAR(multi-scale autoregressive)模型和各向异性扩散偏微分方程PM模型的声纳图像滤波去噪方法MAR-PM。该方法首先利用MAR得到多尺度图像序列,然后对每一级图像进行PM滤波,并预测出最细尺度图像的滤波效果,再以此为先验信息指导对原始噪声图像的PM滤波去噪,最后用PSNR值对处理效果进行评价。试验结果表明:该方法不仅有效提高了声纳图像的滤波效果,而且缩短了处理时间。
To solve the problems of the excessive diffusion and the equation aberrance caused by inac- curate parameters which often appear in the use of the traditional PM （Perona-Malik）filtering model to deal with sonar images, this paper proposes an image denoising method called MAR-PM in combi- nation with partial differential equation and MAR （Multi-scale AutoRegressive）. First, MAR is used to obtain a series of multi-scale images. Then, the PM method is adopted to filter the images on every scale so as to predict the finest-scale image. Using the predicted image as priori information, it is pos- sible to denoise the original image by PM filtering. Finally, the PSNR value is used to evaluate the effect of the method. The experiment result indicates that the method can eliminate the sonar image noises better in a shorter time than any other traditional methods, so that it will have a good prospect of engineering application.
出处
《海军工程大学学报》
CAS
北大核心
2013年第6期53-57,共5页
Journal of Naval University of Engineering
基金
高等学校博士后专项科研基金资助项目(20090641460)
