A Modified Multiplicative Intrinsic Component and Local Intensity Inhomogeneity for Brain Image Segmentation

Abstract

Intensity inhomogeneity is common in medical images and inevitably leads to many challenges for accurate image segmentation. Image segmentation is a foundation step for computer vision and computing. This paper proposes an energy minimization method based on fuzzy C-means (FCM), which combines global with local clustering method for bias field estimation and segmentation of magnetic resonance (MR) brain images. The proposed method takes full advantage of the decomposition of MR images into two multiplicative components as the global clustering term, which uses the true image that characterizes a physical property of the tissues and the bias field that accounts for the intensity inhomogeneity, and their respective spatial properties. The decomposition of MR images describes the varies of bias field of the whole image, where the details of some deep changes of tissue boundary may lose. And the local clustering term which uses varies of bias field in local regions of image for the proposed method can well deal with the deep change of intensity between different tissues. Because of there is a lack of global control for distribution of bias field in local clustering method, we use the advantages of global and local clustering, and consider the combination of them. In the proposed method, bias field estimation and tissue segmentation are simultaneously achieved by an energy minimization process. The energy minimization is iteratively optimized by FCM. Comparison experiments on some real and synthetic images with relevant models demonstrate the advantages of the proposed model in terms of bias correction and segmentation accuracy.