小波多分辩率分析在图像滤波中的应用

The Application of Wavelet Multi-resolution Analysis in Image Filtering

小波闲值滤波基本思想是在小波分解后的各层系数中,对模大于和小于某个阈值的系数分别进行处理。主要针对图像滤波,利用正交小波(db4)和双正交(9/7小波)对三种滤波函数尺度的相关性和滤波图形效果进行比较。由实验可以发现,正交小波(db4)时,半软阈值函数滤波效果较好;双正交小,波(9/7小波)软阚值函数滤波效果较好;正交小波(db4)相对于双正交小波(9/7小波)有效的滤波特性。最后,还通过中值滤波算法对硬阈值滤波函数进行改进。

The main idea of wavelet threshold filtering is based on the processing method of the wavelet coefficients. which is bigger or smaller than threshold coefficient. The processing methnd is according to the absolute figures. The article is aimed at filtering noise in image with orthogonai wavelet （db4） and biorthogonai wavelet （9/7） in three filtering functions. The consequences are compared in two aspects：one is relativity of scalevalues,another is effects of filered images. According to the experiments,we can conclude that half-soft threshoht filtering fimction does better with orthogonal wavelet （db4） and soft threshold filtering function with biorthogonal wavelet （9/7） ,and orthogonal wavelet （dh4） has better filtering specialities than biorthogonal wavelet （9/7）. In the end, the hard threshold filtering function is amcliorated with median filtering algorithm.