【摘要】 设计了一种脉冲耦合神经网络(PCNN)结合最小交叉熵理论的图像二值分割算法,可提高监管码识别过程中字符图像二值分割的准确度,并适合于在FPGA上实现.根据不同灰度图像点火时刻不同的特性,使用PCNN算法对图像进行迭代处理,并计算每次处理后的交叉熵.再利用最佳分割图像与原图像的交叉熵最小的理论,确定最佳迭代处理次数,从而得到最佳的二值分割图像.该方法实现了图像的自动分割,效果由于传统的大津(OSTU)算法.提出了该算法在FPGA上实现方案,实验结果验证了方案的可行性.更多还原

【Abstract】 A bstract An algorithm that combines pulse coupled neural network( PCNN) and minimum cross entropy theory is proposed, which can improve the accuracy of the binary segmentation of character images in the regulatory code recognition and suitable for implementation on field programmable gate array( FPGA). During the image segmentation processing with the PCNN algorithm, different gray pixel will be ignited at distinct iteration time. The image segmentation procedure is performed recursively using the PCNN algorithm, and the cross entropy of each segmented image and the original image is calculated. Based on the theory that the best binary segmentation image has the less cross entropy value with the original image, the best iteration time can be set, and so do the best binary segmentation image. The proposed method can be used for automatic image segmentation, which has better result than the OSTU algorithm. The FPGA implementation theme of the algorithm is proposed and confirmed by the simulation results.更多还原