红外图像的各向异性分段高斯滤波(英文)

Anisotropic Piecewise Gaussian Filter of Infrared Image

提出了一种各向异性分段高斯滤波,这种方法除了考虑图像的尺度和方向特性外,还根据边缘区域可分特性,使用了分段滤波的方法来进一步解决边缘保持的问题.该方法通过本文所提出的一种图像噪音方差估计模型来确定图像的基准尺度,并由Hermite变换得到其方向角,然后再通过确定高斯模型的轴向比和自适应尺度,使对选择区域的滤波转变为对分段高斯模型的滤波,从而使计算的可靠性得到增强.从仿真结果可以看出,各向异性分段高斯滤波器在噪音去除和边缘保持的综合性能上要优于其他常用的滤波算法.

An anisotropic piecewise gaussian filter （PGF） has been putted forward. In this method, not only the information of scale and orientation has been applied, but also piecewise method has been used to preserve edge. In this article, a statistic model of noise has been used to estimate the variance of noise and based on this the standard scale of image has been decided. Then the Hermite transform method has been used to determine orientation adaptively. By calculating the axial ratio and scale of gaussian model, the effective neighbor filter can be converted into piecewise gaussian filter. The experimental result shows that this new method has better capability than other often used method in suppressing noise and preserving edge.