压缩感知的天文图像去噪算法
发布时间：2019-01-09 10:52
【摘要】：针对压缩感知迭代收缩阈值算法在图像处理中存在收敛速度慢和去噪性能差的缺陷,提出了一种改进的高性能迭代收缩阈值天文图像去噪重建算法.首先,使用经典最速下降法中的BB线性搜索步长算子加快迭代收缩阈值算法的收敛速度;其次,为了进一步提高重构天文图像的质量,在传统Visu Shrink收缩阈值的基础上,提出一种下降Visu Shrink收缩阈值对图像信息进行筛选;由于阈值去噪方法在迭代重建的过程中会导致重建的图像中出现伪吉布斯效应,最后采用循环平移的方法在每次迭代过程中对获取的重建图像进行调整.多次的试验结果表明,与传统的压缩感知迭代收缩阈值算法相比,所提出的算法不仅能够获得较优的去噪性能和较快的收敛速度,同时可以有效地保护天文图像的特征和纹理等细节信息.此外,当选取的压缩采样比较低时,本算法也可以获得相对较高的峰值信噪比和视觉质量,进一步验证了本算法在天文图像去噪中的有效性.
[Abstract]:In order to overcome the shortcomings of slow convergence and poor denoising performance in image processing, an improved iterative shrinkage threshold algorithm for astronomical image denoising and reconstruction with high performance is proposed. Firstly, the BB linear search step operator in the classical steepest descent method is used to accelerate the convergence speed of the iterative shrinkage threshold algorithm. Secondly, in order to further improve the quality of reconstructed astronomical images, based on the traditional Visu Shrink shrinkage threshold, a reduced Visu Shrink shrinkage threshold is proposed to filter the image information. Since the threshold denoising method will lead to pseudo-Gibbs effect in the reconstructed image during iterative reconstruction, the method of cyclic translation is used to adjust the reconstructed image during each iteration. The experimental results show that compared with the traditional compression sensing iterative shrinkage threshold algorithm, the proposed algorithm can not only achieve better denoising performance and faster convergence rate, At the same time, it can effectively protect the features and texture of astronomical images and other details. In addition, when the selected compression sampling is low, the proposed algorithm can obtain relatively high PSNR and visual quality, which further verifies the effectiveness of the proposed algorithm in astronomical image denoising.
【作者单位】： 哈尔滨工业大学控制与仿真中心;
【基金】：国家自然科学基金(61074127)
【分类号】：TP391.41
本文编号：2405526
[Abstract]:In order to overcome the shortcomings of slow convergence and poor denoising performance in image processing, an improved iterative shrinkage threshold algorithm for astronomical image denoising and reconstruction with high performance is proposed. Firstly, the BB linear search step operator in the classical steepest descent method is used to accelerate the convergence speed of the iterative shrinkage threshold algorithm. Secondly, in order to further improve the quality of reconstructed astronomical images, based on the traditional Visu Shrink shrinkage threshold, a reduced Visu Shrink shrinkage threshold is proposed to filter the image information. Since the threshold denoising method will lead to pseudo-Gibbs effect in the reconstructed image during iterative reconstruction, the method of cyclic translation is used to adjust the reconstructed image during each iteration. The experimental results show that compared with the traditional compression sensing iterative shrinkage threshold algorithm, the proposed algorithm can not only achieve better denoising performance and faster convergence rate, At the same time, it can effectively protect the features and texture of astronomical images and other details. In addition, when the selected compression sampling is low, the proposed algorithm can obtain relatively high PSNR and visual quality, which further verifies the effectiveness of the proposed algorithm in astronomical image denoising.
【作者单位】： 哈尔滨工业大学控制与仿真中心;
【基金】：国家自然科学基金(61074127)
【分类号】：TP391.41
【相似文献】
相关期刊论文 前1条
1 张强;魏继增;;天文图像差异算法的SoC实现[J];计算机工程;2012年05期
相关硕士学位论文 前1条
1 李洋;压缩感知在天文图像中的应用研究[D];南京信息工程大学;2014年
，本文编号：2405526
本文链接：https://www.wllwen.com/kejilunwen/ruanjiangongchenglunwen/2405526.html
