摘要
为了提高重建图像质量,减少处理时间,提出一种基于L_(1/2)正则约束的单帧图像超分辨率重建算法.该算法在稀疏重建字典对训练阶段,为了有效提取低分辨率图像边缘、纹理等特征细节信息,采用小波系数单支重构方法对低分辨率图像进行特征提取;而在图像重建阶段,为了解决基于L1正则模型得到的解时常不够稀疏,重建图像质量有待进一步提高的问题,采用L_(1/2)范数代替L1范数构建超分辨率重建模型,并且采用一种快速求解的L_(1/2)正则化算法进行稀疏求解.实验结果表明:与现有算法相比较,该算法在重建图像主观和客观评价指标、算法运行速度等方面均更优.
In order to improve the quality of the reconstructed image and reduce the processing time,a single frame image super-resolution reconstruction algorithm based on L_（1/2） regularization constraint was proposed.At the stage of training sparse reconstruction dictionary,the wavelet coefficients single branch reconstruction method was effectively used to extract the features of low resolution image patches.At the stage of image reconstruction,the L_（1/2） regularization model was adopted to instead of the L1 norm,to solve the problem that the solution obtained by the L1 regularization model was often not sparse enough,and the quality of reconstructed image needed to be improved.A fast algorithm of L_（1/2） regularization for sparse solution was presented.The experimental results show that compared with the existing algorithms,the algorithm is better in the reconstruction of the subjective and objective evaluation of the image indicators,running speed and so on.
出处
《华中科技大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2017年第6期38-42,共5页
Journal of Huazhong University of Science and Technology(Natural Science Edition)
基金
国家自然科学基金资助项目(61363078)
兰州理工大学博士科研基金资助项目
关键词
重建图像
超分辨率
稀疏表示
L(1/2)正则模型
小波系数单支重构
reconstructed image
super-resolution
sparse representation
L（1/2） regularization model
single branch reconstruction of wavelet coefficients
