液晶与显示, 2017, 32 (11): 905, 网络出版: 2017-12-01
基于小波域Curvelet变换的湍流图像去噪算法
Atmospheric turbulence image denoising algorithm based on wavelet-domain curvelet transform
图像处理 图像去噪 Curvelet变换 大气湍流 阈值 image processing image denoising curvelet transform atmospheric turbulence threshold value
摘要
为了提高湍流图像的空间分辨率, 提出了一种基于小波域Curvelet变换(wavelet domain Curvelet transform, WDCT)的湍流图像去噪算法。该算法根据湍流退化图像噪声的统计特性, 结合贝叶斯萎缩方法优化阈值选择。首先, 对含噪湍流图像进行单层二维离散小波变换, 接着提取高频系数并对它作快速离散Curvelet变换, 最后根据贝叶斯准则估计阈值T, 改进阈值的自适应选取方法, 获得最优阈值, 最后给出湍流图像去噪实现过程。为验证本文算法, 根据客观评价标准峰值信噪比(peak signal to noise ratio, PSNR)和均方根误差(mean square error, MSE), 对模拟图像和实测湍流图像进行去噪实验。与DWT-NABayesShrink算法、UWT算法相比, 视觉效果更好, PSNR值分别提高7.27%和4.92%, MSE值分别降低26.3%和23.1%。本文算法得到较清晰的目标图像, 对湍流退化图像去噪有一定的应用价值。
Abstract
To enhance the spatial resolution of the atmospheric turbulence image, an atmospheric turbulence image denoising algorithm based on wavelet domain Curvelet transform (WDCT) is proposed in this paper. This algorithm bases on the statistical property of the image noise and combines with Bayes Shrink theory to optimize threshold selecting. Firstly, the turbulence degraded image is performed to a single 2-D discrete wavelet transform (2-D DWT), then extracts the high frequency coefficients and make the fast discrete Curvelet transform for the degraded image. Finally, we estimate the threshold value T according to the Bayesian criterion, and improve the adaptive method of selecting threshold, obtain the optimized threshold. Therefore, the implementation process of the proposed algorithm is addressed. In order to verify the effectiveness of the proposed denoising method, basing on the objective evaluation that are the peak signal to noise ratio (PSNR) and mean square error (MSE), a series of denoising experiments are performed on simulated images and practical observed turbulence image. The experiment results show that, compared to DWT-NABayesShrink method and UDWT method, the visual effect is better, PSNR value has improved 7.27% and 4.92%, respectively, and MSE value are degraded 26.3% and 23.1%, respectively. Our algorithm can obtain the clear image, so the research results have application values for turbulence image denoising work.
王珺楠, 邱欢, 张丽娟, 李阳, 刘颖. 基于小波域Curvelet变换的湍流图像去噪算法[J]. 液晶与显示, 2017, 32(11): 905. WANG Jun-nan, QIU Huan, ZHANG Li-juan, LI Yang, LIU Ying. Atmospheric turbulence image denoising algorithm based on wavelet-domain curvelet transform[J]. Chinese Journal of Liquid Crystals and Displays, 2017, 32(11): 905.