结合最大方差比准则和PCNN模型的图像分割

Image segmentation with PCNN model and maximum of variance ratio

脉冲耦合神经网络(PCNN)模型在图像分割方面有着很好的应用。在各项参数确定的情况下,其分割结果的好坏取决于循环迭代次数的多少,而PCNN模型自身无法实现迭代次数的自动判定。为此提出一种结合最大方差比准则的PCNN迭代次数自动判定算法,用于实现图像的自动分割。算法利用最大方差比准则找到图像的最优分割界限,确定PCNN的迭代次数,获得最优图像分割结果,然后利用最大香农熵准则验证分割结果。实验表明:提出的算法实现了PCNN迭代次数的自动判定,提高了PCNN的迭代速度,运行效率优于基于2D-OTSU和基于交叉熵的自动分割算法,图像分割效果良好。

The Pulse Coupled Neural Network （PCNN） model is very suitable for image segmentation. With given parameters, the results of segmentation are determined only by the times of iteration. However, the PCNN model itself cannot automatically discover the optimal iteration times. Therefore, an algorithm based on the maximization of variance ratio criteria is proposed to solve this problem. The algorithm can automatically discover the best iteration times by applying the maximization of variance ratio criteria, and get the best segmentation results. Eventually, the Shannon entropy rule is used to check the segmentation results. The experimental results show that the algorithm can automatically discover the optimal iteration times, the segmentation results are satisfactory, and it improves the speed of PCNN iteration, and it is also more efficient than the automatic segmentation algorithm based 2D-OTSU and cross-entropy.