图像阈值去噪中的小波特性分析及算法改进

选取合理的小波和阈值算法能够获得理想的图像阈值去噪结果；分析了小波正交性、消失矩以及滤波器结构等特性对图像阈值去噪的影响，提出了图像闽值去噪中小波特性的选取依据．按照噪声与图像信息在小波多尺度分解中传递特性的不同，提出基于信息量的自适应分层阈值算法；仿真结果表明，在进行图像阈值去噪时应优先选取双正交小波，且其消失矩阶数近似于图像的最高次奇异性，去噪小波的高通分解滤波器应具有偶对称性质，低通分解滤波器长度不宜过短．自适应分层阈值算法的图像去噪效果要明显优于全局阈值算法，图像信噪比能够提升至27．86（bior2．6，SNR0=20）；具有偶对称高通分解滤波器的双正交小波能够获得较好的图像去噪效果，其消失矩阶数应近似等于图像的最高次奇异性，基于信息量的小波自适应阈值算法在提升图像信噪比的同时，能够有效保留图像的细节信息．

Analysis of Wavelet Characters in Image Threshold Denoising and Improvement of Algorithm

The effect of wavelet characters in image threshold denoising and analysis is studied. The algorithm of choosing threshold is analyzed and improved. For the particularity of image threshold denoising, the orthogonal, vanishing moments and the filter construction of wavelet in image denoising are analyzed. The necessary condition of wavelet to be chosen is put forward. In accordance with the signal and noise transfer characteristics on multi-scale decomposition, the adaptive stratified threshold algorithm is improved. The simulation results show that the choosing ＂method of wavelet in image denoising is proper and feasible. The denoising image proves the adaptive stratified threshold algorithm based on information content is more effective than global algorithm. The SNR of image using blot2. 6 can be up to 27.86（SNR0 =20）. These results indicate that it is prior to using biorthogonal wavelet in image threshold denoising. The adaptive stratified threshold algorithm is able to improve SNR with more detail information reserved.