基于非均匀五株采样滤波器组的图像阈值除噪

Image Denoising Based on Nonuniform Quincunx Directional Filter Banks via Wavelet Shrinkage

针对传统二维小波缺少多方向特性和传统多方向滤波器组不能进行多尺度分解的缺陷,提出了一种具有多尺度多方向特性的非均匀五株采样滤波器组。该滤波器组每层分解具有一个低频子带和十二个高频子带,并可通过对低频子带的重复分解得到图像的多尺度信息,实现了多方向分解和多尺度分析的统一完成。将提出的非均匀五株采样滤波器组结合统一阈值方法应用于图像除噪。经试验验证,与小波统一阈值除噪相比,所提出的方法对除噪后的图像无论从视觉效果和峰值信噪比(PSNR)上都有较显著提高。

Traditional 2D-wavelet transform lacks multidirectional information. Directional filter banks lacksmultiresolution property. An image denoising method based on nonuniform quincunx directional filter banks via wavelet shrinkage has been proposed. The nonuniform quincunx directional filter banks has twelve highpass subhands and one lowpass subband in each level, and the multiresolution information can be obtained by repeating decomposition in lowpass subband. By wavelet shrinkage method to realize image denoising, the experiment results show that the performance of the proposed method is obviously superior in both vision and Pulse Signal to Noise Ratio （PSNR）.