摘要
对于消除图像中存在的噪声,目前并没有通用的去噪标准。在对图像去噪难点的分析和研究的前提下,简要介绍了该领域最新出现的DUDE(Discrete Universal Denoising)方法,同时对该方法作了改进,重新定义了像素领域,建立较准确的像素纠正方法,然后对受污染的图像进行去噪处理,并且和其他滤波方法作了性能比较。实验结果表明,采用该方法减小了均方误差,降低了内存要求,减少了先验知识或隐含假设,明显地提高了主客观去噪效果,其性能指标非常接近HMM(Hidden Markov Model)。
Considering that there is no universal denoising standard, in order to remove the image noise, under the premise of analysis and research on difficulties in image denoising, it is briefly explained the basic principles of Discrete Universal Denoiser （DUDE） model recently appeared in this field. Meanwhile, some improvements are made for this new algorithm, in which the pixel neighborhoods are redefined to set up a more accurate method to correct pixel intensities. Then the contaminated images will be denoised. Comparisions with the performance of other filtering methods are made. Eexperimental results show that the method not only reduces the mean square error（MSE）, the requirement of memory and prior knowledge or implicit assumption, but also significantly improves the subjective and objective effects of image denoising, and its performance is very close to that of HMM（Hidden Markov Model）.
出处
《半导体光电》
CAS
CSCD
北大核心
2010年第2期296-299,311,共5页
Semiconductor Optoelectronics
