In this paper, a novel context-dependent fuzzy set associated statistical model-based intensity inhomogeneity correction technique for magnetic resonance image (MRI) is proposed. The observed MRI is considered to be affected by intensity inhomogeneity and it is assumed to be a multiplicative quantity. In the proposed scheme the intensity inhomogeneity correction and MRI segmentation is considered as a combined task. The maximum a posteriori probability (MAP) estimation principle is explored to solve this problem. A fuzzy set associated Gibbs’ Markov random field (MRF) is considered to model the spatio-contextual information of an MRI. It is observed that the MAP estimate of the MRF model does not yield good results with any local searching strategy, as it gets trapped to local optimum. Hence, we have exploited the advantage of variable neighborhood searching (VNS)-based iterative global convergence criterion for MRF-MAP estimation. The effectiveness of the proposed scheme is established by testing it on different MRIs. Three performance evaluation measures are considered to evaluate the performance of the proposed scheme against existing state-of-the-art techniques. The simulation results establish the effectiveness of the proposed technique.

中文翻译：

---

用于从磁共振图像进行强度不均匀性校正的上下文相关模糊关联统计模型

在本文中，提出了一种新的上下文相关模糊集关联的基于统计模型的磁共振图像（MRI）强度不均匀性校正技术。观察到的 MRI 被认为受强度不均匀性的影响，它被假定为一个乘数。在所提出的方案中，强度不均匀性校正和 MRI 分割被认为是一个组合任务。探索了最大后验概率（MAP）估计原理来解决这个问题。与吉布斯马尔可夫随机场 (MRF) 相关的模糊集被认为是对 MRI 的空间上下文信息进行建模。观察到 MRF 模型的 MAP 估计在任何局部搜索策略下都不会产生好的结果，因为它会陷入局部最优。因此，我们利用基于变量邻域搜索 (VNS) 的迭代全局收敛准则的优势进行 MRF-MAP 估计。所提出方案的有效性是通过在不同的 MRI 上对其进行测试来确定的。考虑了三种性能评估措施来评估所提出方案相对于现有最先进技术的性能。仿真结果证明了所提出技术的有效性。