基于随机采样的快速非局域均值图像去噪算法

在图像去噪的过程中,针对非局域均值滤波(NLM,non-local mean)去噪方法中存在的运算效率低下问题,提出了一种改进的快速去噪算法.首先,基于蒙特卡罗重要性采样思想,采用Metropolis方法在搜索窗口中对图像的像素点进行随机采样,取代NLM算法在搜索窗口中逐一像素遍历进行加权平均的方法,从而实现了加速.其次,算法在实现加速的过程中一般会影响去噪效果,为了仍保持较好的去噪效果,提出一种改进的权重函数.最后,进行了对比实验,对实验结果从峰值信噪比、运行时间及与原无噪声图像的相似度三个方面进行分析,结果表明,本方法在保持较高信噪比的基础上,处理时间只有标准NLM算法的1/3左右.

Fast Non-Local Means Image Denoising Algorithm on Random Sampling

The volumetric water content of soil is measured by using dielectric property.Choosing the high frequency capacitance method as the measurement method,the water contents of three kinds of arable soil,like red soil,moisture soil and rice soil,are measured.Firstly,the test principle is analyzed,using high frequency electric model of testing plate of soil to connect with fixed inductor in series for getting the relationships between the equivalent capacitance of testing plate and series connection resonant frequency.Then,the theoretical relation curve of resonant frequency and soil volumetric water content is drawn through the process of calculation and circuit simulation.Eventually,by experiment to the water content of arable soil,the comparison chart between the practical measured value and theoretical value is obtained.The actual measurement result conforms the theoretical calculation,which has application value.