基于矩形方向窗的小波域去噪方法

Method of wavelet domain denoising based on directional rectangle window

根据小波变换的特点,提出了一种新的局部邻域窗口选择方法——基于方向性的矩形窗口选择方法。对同一尺度不同子带选用不同方向的矩形窗口,并且不同尺度下窗口的大小也不同;对于多方向性的图像,使用双树复小波变换取代传统的离散小波变换。实验结果表明,将其应用于图像去噪,简单有效,并且可得到更高的峰值信噪比和更好的视觉效果。

According to the features of the wavelet transform, a new method of choosing the local neighborhood was proposed. For the three subbands of the same scale, we chose rectangle windows of different directions, and the size of the rectangle windows for different scales is also different. For the multidirectional images, Dual-Tree Complex Wavelet Transform （DTCWT） was employed instead of the traditional Discrete Wavelet Transform（ DWT）. The experimental results indicate that the method of choosing directional rectangle windows is simple and effective, and high PSNR and good visual quality can be obtained from this method.