频谱包络滤波器及其应用

Spectrum Envelope Filter and Application

针对图像处理中的保边平滑问题,提出了一种新的保边平滑滤波器.滤波器通过分析自然图像的频谱特性以及背景噪声在图像频谱中的分布规律,将图像频谱分为对重构图像边缘起重要作用的能量聚集区和背景噪声区.先用傅里叶变换计算输入图像的频谱,滤波时的滤波器通过模拟自然图像的频谱包络来修正该频谱,这样在保留能量聚集区中的频率的同时抑制背景噪声区中的频率,使得滤波后的图像仍然具有原始图像频谱的显著特征,而修正后的频谱经过傅里叶逆变换可以重构图像.实验结果表明,所提滤波器在平滑背景噪声的同时,能够更好地保护原始图像的边缘信息,保边性能优于传统高斯滤波器,滤波结果的视觉效果也明显优于均值漂移滤波器.

A new edge preserving filter is proposed for the edge protection when the background noise is smoothing during image processing.The filter classifies the image spectrum into two areas by analyzing the spectrum characteristics of natural images and the distribution of background noise in image spectrum：one is the energy-concentrated area that plays an important part in rebuilding the edges;and the other is the background noise area.The spectrum of an imported image is calculated by the Fourier transformation.The spectrum is modified during filtering by the filter which gets correction factors by simulating the spectrum envelope of nature image so that the frequency of energy-concentrated area can be preserved and that of background noise area can be restrained.The filtered image still holds the main characteristic of original spectrum,which can be calculated from the modified spectrum using the inverse Fourier transformation.Experimental results show that the proposed filter can well protect the edge information of original image when the background noise is smoothing,has better performances than the traditional Gaussian filter in edge preserving,and creates a better subjective impression than the mean-shift filter.