摘要
为增强NLSPCA(非局部稀疏主成分分析)算法对去除图像泊松噪声性能,提高图像块聚类精确度,增大字典下的表示系数稀疏性,改善恢复图像易模糊等问题,提出基于L_(1/2)范数的非局部PCA泊松噪声图像恢复改进算法(L_(1/2)-NLSPCA)。新算法首先对图像分割成重叠块;其次采用设计的自适应Bregman K-means算法对分割的图像块聚类;最后使用PCA构建基于L_(1/2)范数的非局部字典下的稀疏表示系数,对聚类后的图像块分组进行去噪重构。实验结果表明,L_(1/2)-NLSPCA算法与基准算法相比峰值信噪比(PSNR)提高了0.52~2.57 dB,在视觉上纹理细节更清晰。
In order to mitigate the issue of image blurring during restoration by using the original NLSPCA(Non-Local Sparse Principal Component Analysis),we propose a novel non-local PCA Poisson noise image restoration algorithm based on L_(1/2) norms(L_(1/2)-NLSPCA)to improve enhance the performance in removing Poisson noise from images.Firstly,the proposed method segments the image into overlapping blocks;secondly,the designed adaptive Bregman K-means algorithm clusters the segmented image blocks to improve the accuracy of image block clustering;finally,we utilize PCA to construct a non-local dictionary and obtain sparse representation coefficients based on L_(1/2) norms,which are subsequently employed in the denoising and reconstruction of the clustered image blocks.L_(1/2) norms can increase the sparsity of the representation coefficients under the dictionary more efficiently.Experimental results show that the L_(1/2)-NLSPCA algorithm improves the peak signal-to-noise ratio(PSNR)by 0.52 to 2.57 dB compared to with the benchmark algorithm,and the texture details are clearer visually visually clearer.
作者
李欢
张文娟
黄姝娟
肖锋
LI Huan;ZHANG Wenjuan;HUANG Shujuan;XIAO Feng(College of Sciences,Xi’an Technological University,Xi’an 710016,Shaanxi,China;College of Computer Science and Information Engineering,Xi’an Technological University,Xi’an 710016,Shaanxi,China)
出处
《咸阳师范学院学报》
2024年第2期10-15,30,共7页
Journal of Xianyang Normal University
基金
国家自然科学基金面上项目(62171361)
陕西省重点研发计划(2022GY-119)
陕西省科技厅自然科学基础研究计划项目(2021JM-440)
陕西省科技厅工业攻关项目(2020GY-066)。
