基于分数阶信息的鲁棒图形模糊聚类分割算法

为改进现有的图形模糊聚类算法不适合图像分割需要的不足,提出像素邻域分数阶信息约束的鲁棒图形模糊聚类分割算法。在现有的图形模糊聚类目标函数基础上,对其聚类所涉及的中立度和拒分度进行正则化约束,采用最优化数学方法推导获得一种图形模糊聚类算法。为增强图形模糊聚类分割算法的抗噪鲁棒性,将像素邻域分数阶积分滤波信息嵌入正则化图形模糊聚类目标函数,通过推导获得适合噪声干扰图像分割需要的鲁棒算法。测试结果表明,该分割算法能够有效降低复杂背景对图像分割目标的影响,提高了邻域均值信息约束的鲁棒分割算法的抗噪能力。

Robust picture fuzzy clustering segmentation algorithm based on fractional information

Aiming at overcoming the shortcomings of the existing picture fuzzy clustering algorithm that it is not suitable for image segmentation, a robust picture fuzzy clustering segmentation algorithm based on the fractional information from the neighborhood of each pixel was proposed. On the basis of the existing fuzzy picture clustering objective function, the regularization item combining neutral degree with refusal degree involved in the picture fuzzy clustering was constructed, and embedded into the objective function of existing picture fuzzy clustering to form the new constrained optimization problems. A picture fuzzy clustering segmentation algorithm was obtained using optimization mathematical method. To enhance the robustness of picture fuzzy clustering segmentation algorithm, the fractional order integral filtering information of pixel neighborhood was embedded into the regularized picture fuzzy clustering objective function, and a robust image segmentation algorithm based on picture fuzzy clustering was induced. The testing results of many remote sensing and medical images show that the proposed segmentation algorithm can not only effectively reduce the influence of complex background on target segmentation, but improve the noise suppression capability of robust picture fuzzy clustering segmentation algorithm constrained by the mean information of neighborhood pixels.