噪声对图像分析、图像压缩的性能有严重的影响.经典的图像滤波器在抑制噪声的同时会丢失图像中的细节.分析了噪声在小波变换下，二进小波变换的极大值在不同尺度下的传播特性之间的关系,并构造了一个一般形式的检测噪声点的四叉树.通过对四叉树的遍历，对图像中椒盐噪声点和高斯噪声的极大值点进行多尺度的定位.实验证明这种方法能有效去除图像噪声，最大限度地保存图像的细节.%Noise has an important effect on image analysis and imagecompression. Classical filters can restrain it with the loss of image details. The transmission characteristic of signal's peak value in multi-resolution wavelet transforms is analyzed and a spatial orientation tree is constructed. Then, the peak value points of the pepper-salt noise or Gauss noise can be located with multi-resolution by going through the hierarchical tree. Experiment shows that this method can restore image with minimum image details loss.
展开▼
