基于梯度倒数的无人机遥感图像融合滤波方法

为了实现对无人机遥感图像去噪处理时兼顾滤波和边缘保持效果,提出了一种基于梯度倒数的无人机遥感图像融合滤波方法——梯度倒数自适应开关均中值滤波。首先计算出模板内中心像素点与邻近像素点的梯度导数;然后选取局部梯度变化阈值,并定义标志数组来标记各像素点是否超过梯度变换阈值。最后,如果当前滤波像素点超过梯度阈值,则采用提出的自适应开关均中值滤波;否则采用梯度倒数加权平滑方法处理。该方法结合了图像的梯度信息,利用图像的连通性和相关性原理以及自适应开关均中值滤波算法去噪的优点,在去除高斯噪声和脉冲噪声同时可以很好地保持图像的边缘和细节信息。实验结果表明:与传统梯度倒数加权平滑方法相比,算法滤波后图像的平均梯度提升3. 16%,MSE下降了约5%,可以有效提升滤波后无人机遥感图像质量,应用价值明显。

A Fusion Filter Method for UAV Remote Sensing Image Based on Gradient Inverse

In order to achieve both filtering and edge-preserving effect when denoising UAV remote sensing image, a fusion filter method for UAV remote sensing image based on gradient inverse, gradient inverse adaptive switch mean and median value filter, is proposed . Firstly, the gradient derivatives of the central pixel and adjacent pixels are calculated. Then, select the local gradient threshold and define a flag array to mark whether each pixel exceeds the gradient threshold. Finally, if the current filtered pixel exceeds the gradient threshold, the adaptive switch mean and median filter proposed in this paper is used, otherwise the gradient inverse smoothing algorithm is used. The method proposed in this paper combines the gradient information of the image, using the connectivity and correlation principle of the image and the advantages of the adaptive switching mean and median filter algorithm to remove the salt and pepper noise and impulse noise while maintaining the edge and detail information of the image. The computer simulation experimental results show that compared with the traditional gradient inverse smoothing algorithm, the algorithm this paper proposed can improve the average gradient of the filtered image by 3.5% and MSE dropped by 5%. It can effectively improve the image quality of the filtered UAV remote sensing image, the application value of this algorithm is obvious.