论文部分内容阅读
文中首先从概率论角度对自然图像中的相似现象给出数学描述,进而得到主要结果之一——相似原理,它揭示了非局部加权平均算法的数学原理.然后依据该原理设计一种新的适合于高斯噪音和一致脉冲噪音之混合噪音的滤波方法,简称“MNF算法”.实验结果表明,MNF算法不但在去除图像中混合噪音时其效果明显优于最近Garnett等人提出的Trilateral滤波方法及已有的其他方法,而且在去除脉冲噪音或高斯噪音时其效果亦可与已有之先进方法媲美.此外,文中对MNF算法的控制参数建议了简单计算公式,并使参数个数达到最小因而更便于应用.
First of all, this paper gives a mathematical description of the similar phenomena in natural images from the perspective of probability theory, and then obtains one of the main results - the similar principle, which reveals the mathematical principle of non-local weighted average algorithm, and then designs a new MNF algorithm ". The experimental results show that the MNF algorithm is better than the Trilateral filter proposed by Garnett et al. Recently when the mixed noise is removed from the image Method and other existing methods, and its effect can be compared with the existing advanced methods in removing impulse noise or Gaussian noise.In addition, a simple calculation formula is proposed for the control parameters of the MNF algorithm and the number of parameters is reached The smallest and therefore easier to apply.