英语翻译III.GB-FAB ANISOTROPIC FILTERINGThe adaptive filtering technique that has been implementedfor preprocessing the input image is an improvement of theforward and backward anisotropic diffusion (also called FAB[25],[27]).The FAB anisotropic

英语翻译III.GB-FAB ANISOTROPIC FILTERINGThe adaptive filtering technique that has been implementedfor preprocessing the input image is an improvement of theforward and backward anisotropic diffusion (also called FAB[25],[27]).The FAB anisotropic
英语翻译
III.GB-FAB ANISOTROPIC FILTERING
The adaptive filtering technique that has been implemented
for preprocessing the input image is an improvement of the
forward and backward anisotropic diffusion (also called FAB
[25],[27]).The FAB anisotropic diffusion is a nonlinear feature
preserving smoothing technique that efficiently eliminates the
image noise and weak textures from the image while preserving
the edge information.
The original anisotropic diffusion filtering (PM) has been proposed
by Perona and Malik [26] where they formulated the
smoothing as a diffusive process that is performed within the
image regions and suppressed at the regions boundaries.In order
to achieve this behavior,they developed a mathematical framework
where the central part is played by a diffusion function that
controls the level of smoothing
(1)
where is the image data,is the gradient operator
at the position at iteration ,represents the
diffusion function and div is the divergence operator.The function
is usually implemented using an exponential function
as illustrated in (2),where the parameter controls the level of
smoothing
(2)
It can be observed that the diffusion function is bounded
in the interval and decays with the increase of the gradient
value .The PM filtering is an efficient feature preserving
smoothing strategy,but it has stability problems caused
by the offsets between the input and output image.Another
problem associated with the standard PM filtering is that the
diffusion function acts aggressively upon medium gradients
[see Fig.2(a)] which results in the attenuation of medium edges
in the smoothed image.To eliminate the limitations associated
with the original PM formulation,the FAB anisotropic diffusion
has been proposed [25].The goal of the FAB diffusion
function is to highlight the medium and large gradients that are
noise independent and this is achieved by reversing the diffusion
process.This can be implemented by applying two diffusions simultaneously:
the forward diffusion that acts upon the low gradients
that are usually caused by noise,while the backward diffusion
is applied to reverse the diffusion process when dealing with medium gradients.This can be observed in Fig.2(a) where
it is shown that the FAB diffusion function [defined in (3)] becomes
negative for medium values of the gradient.Nonetheless,
since the function is defined by two parameters the
problems associated with stability are more difficult to control.
To address these problems,Smolka and Plataniotis [27] proposed
the inclusion of a time dependent cooling procedure [see
Fig.2(b)] where the values of the diffusion parameters are progressively
reduced with the increase in the number of iterations
[see (4)]

英语翻译III.GB-FAB ANISOTROPIC FILTERINGThe adaptive filtering technique that has been implementedfor preprocessing the input image is an improvement of theforward and backward anisotropic diffusion (also called FAB[25],[27]).The FAB anisotropic
三.国标- Fab的各向异性过滤自适应滤波技术,已实现输入图像的预处理是一种落后的提出改进和各向异性扩散（也为Fab [25],[27]）.巴西空军各向异性扩散是一个非线性特征,有效地维护平滑技术消除了图像的图像噪声和纹理,同时保留弱边缘信息.原来的各向异性扩散滤波（下午）已经提出了佩罗娜和马利克[26],他们制定了一个扩散的过程,是图像区域内进行,并在边界地区的镇压平滑.为了实现这一目标的行为,他们开发出一种数学框架,中部是一个扩散的作用,它控制平滑（1）的水平发挥的图像数据在哪里,是在梯度算子在迭代的位置,代表的扩散功能分区是发散运营商.该功能通常是用指数函数实现如下图中（2）,其中参数控制平滑（2）可以看出,扩散函数在区间和渐变的价值增加衰变界的水平.项目经理是一个高效的过滤功能,维护平滑策略,但它的输入和输出之间的偏移所造成的图像稳定性问题.下午滤波与标准相关的另一个问题是,积极的扩散功能作用后[见图介质梯度. 2（a）条]这导致了在图像中边缘平滑的衰减.为了消除与原来的下午制订相关的限制,巴西空军已经提出了各向异性扩散[25].巴西空军的扩散函数的目的是突出大中型梯度噪声是独立的,这是由扩散进程中取得扭转.这可以实现同时采用两种扩散：向前扩散之物,通常是由噪音造成的低梯度行为,而落后的扩散应用于扭转时,扩散过程中梯度问题.这可以观察到图. 2（a）如它表明,巴西空军扩散函数变成为负的梯度媒体价值观[在（3）定义].不过,自定义函数是由两个参数与稳定相关的问题更难以控制.为了解决这些问题,Smolka和Plataniotis [27]提出了一个随时间变化的冷却过程[见图列入. 2款（b）]那里的扩散参数的值是在逐步与迭代次数的增加而降低[见（4）]
耨,很累的!