提出一种适用于去除高密度椒盐噪声的图像滤波算法,以进一步提高输出图像的峰值信噪比。利用直方图形状判定椒盐噪声的两种灰度值,用于噪声像素的检测与定位。对于非噪声像素,直接输出灰度值;对于噪声像素,沿其邻域的k个方向分别搜索一个距离最近的非噪声像素,然后以欧式距离倒数为权重,采用k个非噪声像素的加权灰度均值作为噪声像素的输出灰度值。测试了不同的方向数k对滤波性能的影响,确定了k的最佳取值为4。采用该方法对椒盐噪声密度为10%到90%的图像进行滤波,输出图像的峰值信噪比比现有同类方法提高了1.8~4.7 d B。该方法有效提高了高密度椒盐噪声图像的滤波质量,处理速度满足实时要求。
在无人机和地面站通信交互的过程中,由于各方面因素,例如频率不同步、传输延时等,可能会造成无人机采集到的数据在传输期间发生错误,导致地面站接收到的数据有部分丢失。文章提出一种梯度下降优化算法--梯度下降自适应学习率算法(RMSProp with NAG,RMSPN),对缺失数据集进行曲线拟合,得到丢失数据的近似值,对缺失数据集进行填补。实验结果证明了该方法曲线拟合效果良好,估计值与实际值误差较小,算法可行性高。
This paper proposes a latch that can mitigate SEUs via an error detection circuit.The error detection circuit is hardened by a C-element and a stacked PMOS.In the hold state,a particle strikes the latch or the error detection circuit may cause a fault logic state of the circuit.The error detection circuit can detect the upset node in the latch and the fault output will be corrected.The upset node in the error detection circuit can be corrected by the C-element.The power dissipation and propagation delay of the proposed latch are analyzed by HSPICE simulations.The proposed latch consumes about 77.5%less energy and 33.1%less propagation delay than the triple modular redundancy（TMR）latch.Simulation results demonstrate that the proposed latch can mitigate SEU effectively.
