Study On MRI Image Segmentation Algorithm Based On Improved Watershed And Improved Fuzzy C-means

Medical image segmentation is a very important field of application of image segmentation.Medical image segmentation refers to the use of image segmentation techniques in the segmentation which is between lesions and normal tissue.The effect of the segmentation has great impact on the physician's judgment of the patient's condition.Watershed algorithm and gaussian kernel fuzzy c-means clustering algorithm(KFCM)are two of the more common medical image segmentation algorithms.This thesis is aimed to improve the conventional watershed algorithm and the conventional KFCM algorithm in order to use the improved algorithm in MRI brain segmentation.The MRI image with noise must be filtered before we use watershed algorithm and KFCM algorithm.First of all,?r is a key parameter of bilateral filtering.Bilateral filtering is discussed with different ?r.Secondly,the filtering effect of soft threshold of wavelet and hard threshold of wavelet is discussed.Finally,improved bilateral filtering,non-local means filtering and soft threshold of wavelet are used to filter the same noisy MRI brain image.Experiments show that when ?r=0.05,we can obtain ideal filtering effect by using bilateral filtering.The filtering effect of soft threshold of wavelet is better than hard threshold of wavelet.The filtering effect of bilateral filtering is the best of above-mentioned three filtering methods.So bilateral filtering is used for the image pretreatment.The conventional watershed algorithm uses conventional morphological gradient to obtain a gradient image as the watershed transformation input.It can easily cause the problem of over-segmentation.Structuring element in the conventional morphological gradient is sensitive to its size.That is to say,with the structuring element's size increases,the thickness of the edge also increases gradually.To solve this problem,multi-scale morphological gradient is used to replace the conventional morphological gradient in the selection of structuring element of the same shape.It is possible to select larger structuring element to better suppress noise,while not cause gradient image obtained edge thickness increases.The further procedure of the detection of similar regions and the combination of small regions is added to get the final segmentation result.The conventional watershed algorithm and the improved watershed algorithm are used to deal with the same MRI brain image.Experiments show that improved watershed algorithm can effectively reduce over-segmentation,obtain ideal segmentation results.Due to the conventional KFCM algorithm's poor performance on the removal of artificial boundaries,the conventional KFCM algorithm is improved with the combination of gray threshold processing and gray-opening operation.The segmentation effect of the conventional KFCM algorithm and the improved KFCM algorithm is discussed in this thesis.Experiments show that the segmentation result of improved KFCM algorithm is closer to the results of manual segmentation.It shows the effectiveness of the improved algorithm.