基于GPU并行计算的图像去噪研究

Research on Image Denoising Based on GPU Parallel Computing

为更好保留经图像去噪后的特征信息,提高图像去噪计算速率,结合非局部均值算法的原理,提出了一种基于GPU并行计算的图像去噪方法。为了更好的保存去噪后的图像信息,采用欧氏距离与邻域相关系数结合方法,对邻域间相似性进行衡量;为提高算法执行速率,将图像复制到GPU内存,将计算结果复制到主机,从而节约主机开销。对添加不同噪声水平的噪声图像进行测试,结果表明,提出的算法在图像相干斑抑制和计算速率方面都有明显的优势。

In order to better preserve the image denoising feature information, improve the image denoising algorithm, combined with the principle of nonlocal means algorithm, an image denoising method based on GPU parallel computing is proposed. For better preservation of the denoised image information Euclidean distance and neighborhood correlation coefficient method are used to measure of similarity between fields. Secondly in order to improve the algorithm executing rate, the image is copied to the GPU memory, then the results will be copied to the host machine, thus the main overhead can be saved. Finally, the images with different noise levels are tested. The results show that the proposed algorithm has obvious advantages in image speckle reduction and computation speed.