结合灰度熵变换的PCNN小目标图像检测新方法

为了自动地进行小目标图像分割检测,从含单一弱小目标图像的特征出发,提出了一种结合灰度熵变换的脉冲耦合神经网络(PCNN)小目标图像分割检测新方法. 该方法在对有随机噪声和复杂背景图像进行非线性灰度熵变换滤波的基础上,考虑灰度熵值灰度图在满足先验概率目标背景比条件下,选择包含单一小目标局部窗口作为处理图像区域,并在局部最小交叉熵判据下,进行改进型PCNN迭代分割检测处理. 实验结果表明,该方法不仅能可靠地检测出复杂背景及随机噪声干扰下弱小目标,并且在PCNN运行处理过程中,可自动地完成最佳分割检测.

Novel Detection Method Using PCNN Combined with Gray Scale Entropy Transform in Small Target Images

In order to conduct small target image segmentation automatically, a new method based on pulse couple neural networks(PCNN) and the gray scale entropy, is proposed for image segmentation and detection, starting from the aspect of characteristics of single small target image. Based on nonlinear gray scale entropy transform on an image with complex background and stochastic noise, this algorithm takes into account the condition that the gray scale images of gray scale entropy satisfy the object to background ratio of prior probability, and select the local region including a single small target which can be regarded as image processing part. Iterative segmentation and detection using improved PCNN is utilized under the criterion of local minimum cross-entropy. The experimental results show that the novel method not only can detect small target with the disturbance of complex background and random noise reliably, but also implement the best segmentation and detection automatically. This algorithm has stronger adaptability and performs well in target detecting.