Empirical Mode Decomposition And Its Application In Medical Image Processing

Because of the medical images have some characteristics in nature, such as illegibility, inhomogeneity and inferior noise robust, so it become particularly important how to carry through post-processing for medical images which we had obtained, and acquire more valuable medical information. In this paper, A new time-frequency analysis method was studied, empirical mode decomposition method, which was used for medical images'post-processing.The empirical mode decomposition (EMD) was introduced as a new signal processing method by Huang. It is suitable for non-stationary signal processing. The empirical mode decomposition, which is independent of Fourier transform, is a fully data driven method with multiscale features. Since the EMD method has good effects in one-dimensional signal processing, many scholars extend it to bidimensional. So bidimensional empirical mode decomposition (BEMD) methods were introduced for processing bidimensional signal. Because of the complexity of bidimensional signal, the BEMD method has many defects need to be improved.In this paper, the foregone BEMD methods were deeply researched. Aiming at correlative questions, the corresponding resolvents were introduced, and then an improved bidimensional empirical mode decomposition method which was used for medical images'denosing and segmentation's research was proposed.Aming at the chracateristics of MRI medical images'low signal noise ratio, two kinds of medical images denosing method based on BEMD was studied in this paper, the one was medical images denosing based on intrinsic mode function's weighted threshold. After the image was decomposed by BEMD method, it was concerned that the image's noise mainly distribute in the high frequency and intermediate frequency, so the intrinsic mode function's weighted thresholding was applied for denoising. Another was medical images denoising based on BEMD and wavelet thresholding. It was concerned that the wavelet thresholding was used for denoising of the intrinsic mode function's high frequency and intermediate frequency, finally reconstructing the original image to achieve the effect of denoising.Aming at the chracateristics of MRI medical images'organization structure, a medical images segmentation method based on BEMD and Gray-level co-occurrence matrix was studied in this paper. During the medical images segmentation, it was concerned that the smoothness of data will influence the segmentation quality, so during structing the improved BEMD method, an interpolation method base on the conjugate gradient method to optimize the CSRBF was used for construct the envelope. After the image was decomposed by the improved BEMD, the intrinsic mode function distributed from high-frequency to low-frequency was obtained. Then the method that gray-level co-occurrence matrix extracted the pixel characteristics was applied, and KFCM was used for cluster analysis of eigenvector set we obtained, then a coarse segmentation results of medical images were obtained, finally for careful partition. Thereby, it was achieved the method's effective application in medical image segmentation.