基于BP神经网络评价的图像闭环控制迭代分割算法

A Closed-Loop Controlling Iterative Image Segmentation Algorithm Based on BP Neural-Network Evaluation

论文在分析闭环控制分割原理与OTSU算法基础上提出了一种新的分割算法。通过监督训练BP网络实现了局部分割结果评价,并根据评价结果对每个目标区域进行了平滑与擦除处理,同时增加先验目标点出现的最小最大频度参数,实现OTSU算法的闭环参数控制,然后引入迭代机制实现图像的综合分割。最后论文将算法用于显微镜细胞图像分割,试验表明该算法可以大大改善分割效果。

The paper advanced a new segmentation algorithm basing on introducing the principle of closed-loop control segmenta- tion and OTSU algorithm. The paper achieved local segmentation effect evaluation by superstending training BP neural-network, and made smoothing and erasing to every object region through the evaluated results, and added the parameters of minimum and maximal probability of considered points, accomplishing the closed-loop parameter control to OTSU, then realized image synthetical segmentation by iteratire system. Finaly,the methodlogy was used in the cell image segmentation under microscope, and the results provided that the accuracy of segmentation was increased.