一种基于方向梯度的图像滤波算法

A Method of Image Filter Based on Orientation Gradient

自然图像中的大部分图像都存在噪声,在滤除噪声的同时尽可能地保存图像细节是图像平滑的目标.目前普遍使用的均值滤波滤除噪声效果不够理想且易使图像变得模糊,中值滤波虽然滤除噪声效果较好但使得图像变得模糊且易使图像的细节信息丢失.针对这一问题,本文提出了一种基于方向梯度的图像滤波算法.该算法对含有噪声的图像取3×3邻域,构造4个方向的梯度算子模板,通过比较4个方向梯度大小,对噪声、边缘、图像内部点分别处理,实验结果表明该算法在滤除噪声的同时,很好地保存了图像的原始信息且有较好的信噪比.

Natural images are prone to exist a variety of noises. The object of image filter is to remove noises as well as save details simultaneously. And it is easy to make image become fuzzy by using usual averaging filter, the effect of averaging filter is not ideal. Although it has good effect by using median filter, it is easy to make image become mistiness and lose detail information. In allusion to this question, in this paper, we proposed a method of image filter based on direction gradient. This new method selects a 3×3 neighborhood, constructs 4 templets of orientation gradient, and processes noises, edge and inside pixel of image according to the magnitude of 4 templets of orientation gradient. The result of experiment proves that the new method not only saves the original image information and has a good Signal-to-Noise Ratio, but also removes noises simultaneously.