多结构元形态学和PCNN的图像凹点检测

Image Foveation Points Detection Research Based on Multi-Structure Elements Morphology and Pulse Coupled Neural Network

脉冲耦合神经网络(PCNN)模拟人眼视觉机制进行凹点检测时,受背景和噪声影响较大,不易检测出有效凹点。提出一种基于多结构元形态学和PCNN改进模型的凹点检测方法。首先利用多结构元形态学边缘检测器抑制背景和噪声影响,保持图像的层次性和主要细节,然后通过PCNN改进模型进行凹点检测,最后选择适当的阈值调整凹点数目。实验结果表明,凹点的选择准确有效,非常符合人眼视觉机制。

Pulse Coupled Neural Network（PCNN） is used to simulate the human visual mechanism to detect image foveation points,because of the great influence of the background and noise,it is not easy to obtain the effective foveation points.A detection method of image foveation points is proposed,which is based on a multi-structure elements morphology and the improvement model of PCNN.First,the multistructure elements morphology edge detector is applied in the image to restrain the background and noise, preserving the hierarchy and main details of image,then foveation points in the image is detected through the improvement model of PCNN.finally the number of foveation points is adjusted by the suitable threshold.The experimental results indicated that foveation points selected is accurate and effective, conforms to the human visual mechanism.