论文部分内容阅读
为解决用传统粒子群算法估计磁共振(MR)图像偏移场会陷入局部最优的问题,本文提出了一种自适应权重粒子群算法估计MR图像的偏移场。针对传统粒子群算法的缺陷,设计一个衡量早熟收敛程度的指标,根据此指标来自适应地调整惯性权重,确保粒子群有效地进行全局寻优,避免陷入局部最优。本文利用Legendre多项式来拟合偏移场,多项式参数利用本文提出的算法进行寻优,最后对MR图像的偏移场进行估计和矫正。将本文算法与改进的熵最小方法进行对比分析,本文矫正后图像熵值更小,对偏移场估计更准确,将矫正后的图像进行分割,分割精度提高将近10%。研究结果初步说明,本算法可应用于MR图像偏移场的矫正。
In order to solve the problem that traditional particle swarm optimization (PSO) is used to estimate the magnetic resonance (MR) image deflection field will fall into the local optimum, an adaptive weighted particle swarm optimization algorithm is proposed to estimate the MR image offset field. Aiming at the defects of traditional particle swarm optimization algorithm, an index to measure the premature convergence is designed. According to this index, the weight of inertia is adaptively adjusted to ensure that the particle swarm optimization is globally effective and avoid falling into the local optimum. In this paper, the Legendre polynomial is used to fit the offset field. The polynomial parameters are optimized by the proposed algorithm. Finally, the offset field of the MR image is estimated and corrected. Comparing the proposed algorithm with the improved entropy minimization method, the image entropy of the proposed method is smaller and the estimated offset field is more accurate. After the corrected image is segmented, the segmentation accuracy is improved by nearly 10%. The preliminary results show that this algorithm can be applied to the correction of MR image offset field.