论文部分内容阅读
文章提出了一种基于视觉模型的图像边缘检测算法．作者采用具有反馈和前项控制连接的侧抑制模型进行图像信息进行处理，用零阶和二阶厄米特（Ｈｅｒｍｉｔｅ）函数的组合产生系统的控制模板，组合系数通过梯度下降学习算法确定．这种以生物视觉感知机理为基础的神经计算方法可有效地均衡滤除噪声和增强边缘，同时将系统高维参数 的学习问题转化为对少数几个参数的确定，降低了问题的维数和计算的复杂性．
This paper presents an image edge detection algorithm based on visual model. The authors use the side suppression model with feedback and control connections to process the image information. The control template of the system is generated by the combination of zero order and second order Hermite functions. The combination coefficient is determined by the gradient descent learning algorithm. This method of neural computation based on biosensory perception mechanism can effectively filter out noise and enhance the edge. At the same time, the learning problems of high-dimensional parameters of the system are transformed into the determination of a few parameters and the dimension of the problem is reduced Computational complexity.