基于模态振幅的模态单元滤波及其在高斯消噪中的应用

he oscillating cell between two neighbor zerocross points in the intrinsic mode function is defined as mode cell and regarded as the fundamental analysis object, while the idea of traditional threshold denoising method, which regards the isolated point as the threshold object, is disclaimed. The conception of mode amplitude is introduced to identify the types of mode cell by thresholding. Furthermore, a novel mode-cell-oriented mode cell filter (MCF) model based on empirical mode decomposition is proposed. The evolution rules of the mode amplitude of mode cell in cases of decomposing fractional Gaussian noises of different Hurst parameter are analyzed, based on which, the threshold selecting methods for Gaussian denoising are established. Finally, the experiments on Gaussian noise with different Hurst parameter reveal that the performance of the proposed algorithm is overall better than that of optimal wavelet denoising algorithm. The results demonstrate that MCF is a self-adaptive and general-applicable denoising method for color Gaussian noise.