【摘要】 提出了一种基于双谱重构的图像去噪方法 .该方法的基本理论依据是 :高阶统计量对高斯噪声的不敏感性 .它首先计算图像的高阶谱 ,而得到的高阶谱中理论上将不再包含高斯噪声 ,然后再通过高阶谱重构图像的幅值和相位信息 ,最后还原至空间域而得到去噪后的图像 .它是高阶谱域的变换与重构方法 .仿真结果和分析表明 ,该双谱重构法较好地解决了去除噪声与保留细节之间的矛盾 ,无论从直观视觉和从定量分析结果 ,该双谱重构法都优于中值滤波法和高阶累积量方法更多还原

【Abstract】 A novel image denoising algorithm based on the bi-spectral reconstruction was proposed. The basic idea of this new algorithm was: higher order statistics was not sensitive to the Gaussian noise. First of all, this new algorithm would calculate the higher order spectrum of the image, and theoretically this higher order spectrum would not include the Gaussian noise any more. Secondly, it would reconstruct the amplitude and phase information of this processed image by the higher order spectrum. Finally, it would get the denoised image by transforming to the spatial domain. This new algorithm was the algorithm of transformation and reconstruction in higher order spectrum. Simulation results and analysis indicate that this new algorithm could solve the contradiction between reducing noise and reserving the image detail so that in views of both visual quality and quantitative analysis, the new algorithm is better than the median-value filtering algorithm and the higher order cumulant algorithm.更多还原