Segmentation For Brain Mr Image Based On Fuzzy C-Means Clustering And Super-Pixel Algorithm

Medical image segmentation is the basis for three-dimensional reconstruction of human tissues, and it can also provide strong support for clinical diagnosis and adjuvant therapy. As there are electronic noise, offset field distortion and volume effect in the process of magnetic resonance imaging, MR image are always with noise and bias field. Since the resolution is limited to the MRI apparatus, there are volume effects in the MR image. Volume effect is the boundary overlap between different organizations, the boundary pixels is the mixture of various tissues. All these problems increase the difficulty of MR image segmentation, and bright great difficulties and challenges for medical image segmentation.With the development of image segmentation technology, nearly a thousand kinds of image segmentation algorithm is applied to medical image segmentation field at present, and fuzzy clustering algorithm is the most suitable one for medical image segmentation. Fuzzy clustering algorithm can deal with the ambiguity of medical images effectively, and fuzzy C-means clustering algorithm is the most widely used one. As the most classical fuzzy clustering algorithm, there are also some flaws in fuzzy C-means clustering method. In order to correct these deficiencies, scholars have done a lot in the improvement of the clustering algorithm, and this promotes the development and application of fuzzy clustering technique.Super-pixel calculation method is a new segmentation method, it can reduce the computational complexity of image processing effectively when used in the pre-processing stage of image segmentation, which can effectively improve the efficiency of segmentation. Super-pixel can preserve the original boundary information of the image while strengthening local image consistency, and atomic regions obtained by super-pixel contain some characteristics which not exist in single pixel, such like shape, boundary contour and histogram. It is beneficial to improve the accuracy of image processing and reduce the time complexity by using super-pixel. Super-pixel calculation method is widely used in image segmentation preprocessing stage of various fields.Based on the this background, this paper presents a new method named "The multistage medical image segmentation method based on super-pixel and fuzzy clustering", which is also referred to as "MSFCM". The method makes full use of the feature information and spatial information of medical image, and have achieved good results in the actual image segmentation.Firstly, the method divides the image into super-pixels, and then makes deep segmentation on super-pixel which has a larger gray value variance. Secondly, cluster the super-pixels by some certain characteristics using fuzzy C-means clustering algorithm, and obtain the super-pixels’membership matrix. For the super-pixels that lack of clarity of membership, the method can make use of spatial information to determine its classification. Finally, super-pixels that have the same classification should be merged, and the final result can be got. As can be seen from the experimental results, the proposed method effectively overcome the effects of noise and bias field for image segmentation, segmentation accuracy and robustness of the method is significantly better than the fuzzy C-means clustering algorithm.