基于粗糙集和多向差分图像的去噪方法

Denoising Approach Based on Rough Sets Theory and Multi-Directions Difference Image

针对图像滤波方法对噪声类型具有偏好性的不足,提出了一种基于粗糙集的多向差分图像去噪方法.根据多向差分图像提出了描述像素平稳性的方向数和描述噪声性的方向极值数的概念,采用粗糙集理论按照像素的平稳性和噪声性将图像划分为平稳噪声子图、脉冲噪声子图以及正常像素子图.并对两类噪声子图分别设计了不同的自适应滤波算法.通过与其他5种滤波方法的对比实验,结果表明该方法对4种不同噪声在MSE、PSNR和ISNR等性能指标上均获得了最优值,尤其对于椒盐噪声,该方法获得的ISNR指标高出中值滤波35%.

To overeome the drawbaek of the preferenee of image filters for the type of noise, a new denoising approaeh based on rough sets theory and multi-direetions differenee image is proposed. The notions of number of direetions and number of direetional extremums are defined to deseribe the smoothness and noisiness of the pixels aeeording to the multi-direetions differenee image. An image is partitioned into the smooth-noise sub-image, the impulse-noise sub-image and the unpolluted sub-image aceording to the smoothness and noisiness of pixels with rough sets theory. Two adaptive filters are designed for these two kinds of noise sub-images. The experiment results, eompared with the other five filters, shows that the filter designed in the paper is the best in the performance of MSE, PSNR and ISNR for the four types of noise. Especially, for the salt ~pepper noise, the ISNR value of the designed filter is higher than that of the median filter with 35%.