一种基于四分法噪声检测的开关中值滤波算法

Switching Median Filter Algorithm of Noise Detection Based on Quartering

为了精确的检测出图像中的脉冲噪声并滤除,提出一种基于四分法噪声检测的开关中值滤波算法。该算法使用一种新的检测方法将噪声检测窗口内的像素按灰度值大小进行排序,并且通过差分方法划分出高、低阶噪声块和高、低阶信号块四部分。当待测像素属于高、低阶信号块时将其视为信号点,否则,根据噪声块与信号块内像素比例关系就可认定其为噪声点或可能噪声点,若为可能噪声点,则需扩展检测窗口重新检测。对确定噪声点,取滤窗内信号点的中值作为滤波输出。实验证明该算法对脉冲噪声有较强的抑制作用。

To accurately detect and filter out the impulse noise in the image, the switching median filter algorithm of noise detection is proposed based on quartering. The algorithm uses a new detection method to sort pixels by the size of gray value within the window of noise detection, and divided by difference method into high, low noise and high, low signal blocks. When the pixel under test is high, low signal block, it is identified as the signal points; otherwise, according to the pixel ratio in noise and signal blocks, it can be recognized as noise points or possible noise points. If it is the possible noise points, then we need to expand the detection window and re-test. We can determine the noise points through taking the median of the signal points in filter window as the filter output. Experimental results show that the aleorithm can strongly inhibit the imnulse noise.