基于前馈神经网络的图像去噪研究

Research on Image De-noising Based on FFNN

提出一种检测和去除污染图像脉冲噪声的有效算法,采用前馈神经网络(FFNN)检测污染图像脉冲噪声,采用改进的算术均值滤波器去除测出的脉冲噪声。该算法仅对检测到的噪声像素进行滤波,采用窗口中未污染像素的算术平均值去除噪声。实验结果表明,该算法无论在图像质量的定性还是在定量评价方面均获得了良好效果。

A novel efficient algorithm is proposed to detect and remove the impulse noise from the corrupted images.The algorithm uses the feed forward neural network（FFNN） to detect the impulse noise from the corrupted images,and uses the modified the arithmetic mean filter to remove the detected impulse noise.Whenever the detector detects the noisy pixel,that particular pixel alone is filtered.The proposed algorithm restores the corrupted pixel value using the average of the uncorrupted pixels in the selected window.The experimental results show that proposed algorithm produces remarkably good results both in quantitative measures and qualitative judgments of image quality.