基于小波包分解的纹理图像去噪

噪声对图像的后续处理影响较大，常用的去噪方法虽然可以去除变化平缓的图像中的噪声，但对细节较多的纹理图像的去噪效果却不太理想．文中基于信号和噪声在小波分解中呈现出来的不同特性，提出了一种新颖的小波包去噪算法．采用该算法对纹理图像进行最优小波包分解，并计算每个子频带的两个范数，然后根据范数值区分信号和噪声，从而达到去除噪声的目的．实验结果表明，该算法对皮革图像具有较好的去噪效果．不仅可以去除纹理图像中的大部分噪声，而且可以较好地保留图像纹理信息．

Denoising of Texture Images Based on Wavelet Packet Decomposition

Noise greatly affects the post-processing of images. General denoising methods can remove the noise of images with gentle change, but have a poor effectiveness on the denoising of texture images with many details. In this paper, according to the different characteristics that signal and noise exhibit during the wavelet decomposition, a novel denoising algorithm based on the wavelet packet decomposition is presented. In this algorithm, texture images are decomposed by using the best wavelet packet and the two norms of each sub-band are obtained. Thus, signals and noise can be discriminated based on the norms, and the texture images can be denoised. Experimental results indicate that the proposed algorithm is of excellent performance in denoising leather images, and can remove most noise of texture images with well-kept texture information.