基于连续状态小波阈值的各向异性扩散去噪方法

Denoising method of anisotropic diffusion based oncontinuous state wavelet threshold

小波方法和偏微分方程方法是图像去噪中的主要方法。根据二者的去噪特点,提出了一种结合两种方法的混合去噪算法。对噪声图进行小波变换,得到高频子带和低频子带。通过对各高频子带进行归一化,获得一种连续状态量,为了保护边缘对这一连续状态量进行前向-后向扩散。由扩散后的新状态量得到由其决定的权系数,把权系数作用在小波系数上得到去噪后的各高频子带,通过与低频子带的重构得到去噪图像。数值试验结果表明:通过采用本方法对图像去噪,得到了较好的去噪效果,达到了既保护边缘又去除噪声的目的,能获得较高的信噪比。

Wavelet and partial differential equation （PDE） are the main methods in removing image noises. According to both characteristics of denoising, a mixed method of combining wavelet with PDE for image denoising is proposed. Each higher frequency subband is normalized to gain a quantity of continuous states. It is important to pr-otect edge that a continuous state quantity is de noised by means of forward and backward diffusion. We obtain theweight coefficient that is determined by new state quantity after diffusion. The new higher frequency subbands aregained when weight coefficient act on wavelet coefficient. We achieve the process of denoising through wavelet r-econstruction in the end. It is indicated from the experimental results that new method can receive better effect of image de noising and higher SNR than others methods, it attain the purpose of preserving edge and smoothing noi-se at the same time.