基于自回归移动平均模型的图像模糊消除并行稳定机制研究

为了克服图像模糊消除算法不稳定与解模糊等难题，保证复原图像的细节信息清晰完整; 并提高算法的运行效率，获取实时性，本文提出了神经网络融合自回归移动平均模型的图像模糊消除并行稳定机制。引入神经网络，基于突触权重系数，构造激活函数；再嵌入人工蜂群算法（ABC- Artificial Bees Colony），并以神经网络的均方误差函数设计适应度方程，由ABC算法训练神经网络，利用优化后的神经网络来获取自回归移动平均模型的参数；再将自回归移动平均优化模型引入模糊图像，以同时识别模糊函数与模糊图像；并对模糊函数进行相关定义，以消除算法不稳定性与解模糊问题；再对模糊图像进行反卷积，消除模糊。借助仿真实验来测试本文机制的相关性能，结果表明：与其他模糊消除算法相比，该机制的运行速度更快，时耗最短；且本文机制更稳定，模糊消除效果更好，复原图像的细节信息清晰可见。

The Study on Stable Parallel Image Deblurring Mechanism Based on the Weighted Optimized Autoregressive Moving Average Model

In order to overcome the unstable with ambiguity blurring problem of these algorithm, as well as guarantee the clear and complete detail information of restoration image, and improve the computation speed of current image deblurring algorithm to achieve the goal of real time, the real-time stable mechanism for image deblurring based on the autoregressive moving average model. Introducing the neural network, and based on the synaptic weights coefficient to construct the active function; then embedded artificial bee colony algorithm (ABC- Artificial Bees Colony), and designed the fitness function according to the mean square error function of neural network, using the ABC algorithm to train the neural network for finding the optimized weight value of neural network as well as the bias of active function to achieve global minimum; finally, the autoregressive moving average optimized model was designed to simultaneously identify fuzzy functions and fuzzy image to deconvolution the nonlinear deblurring image for eliminating the fuzzy. The performance of this algorithm was tested by the simulation experiments, the results showed that compared with other deblurring algorithm, the running speed of this mechanism is faster, time consuming was the shortest; and the deblurring effect was the best, the detail information of restoration image was clearly visible.