论文部分内容阅读
提出了平面拟合编码的一种新的实现方法，即神经网络方法。为了保证Ｈｏｐｆｉｅｌｄ神经网络的收敛，对该网络模型的迭代算法进行了修改，针对Ｈｏｐｆｉｅｌｄ网络存在的局部极小问题，给出了一种扰动算法，结合初始状态的合理选择，可以有效地避免网络陷入局部极小，而接近全局最小，以求得待定系数的最优解，计算机模拟结果表明，Ｈｏｐｆｉｅｌｄ神经网络实现的平面拟合编码性能优于传统的最小二乘法，重建图像质量提高约０．６ｄＢ。
A new method of planar fitting coding is proposed, that is, neural network method. In order to ensure the convergence of Hopfield neural network, the iterative algorithm of the network model is modified. Aiming at the local minimum problem of Hopfield network, a perturbation algorithm is given. With the reasonable choice of the initial state, the network can be effectively avoided The local minimum and the global minimum are close to get the optimal solution of the coefficients to be determined. The computer simulation results show that the performance of the Hopfield neural network is better than the traditional least square method in the performance of the planar fitting coding, and the reconstruction image quality is improved by about 0.6dB .