摘要
针对图像稀疏重建中因使用固定参数的全变分(TV)正则项所带来的图像细节缺失和阶梯效应问题,提出了一种自适应二阶广义全变分(TGV)约束的图像稀疏重建算法。该算法采用二阶广义全变分模型权衡图像的一阶导数和二阶导数,且能够根据每次迭代得到的重构解及对应张量函数自适应地修正权重系数,实现图像的稀疏重建。与全变分正则模型和固定参数广义全变分正则模型相比,该算法能更好地保持图像轮廓和细节信息,提高重建图像的峰值信噪比(PSNR)和结构相似度(SSIM)。
Consideiing the image detail loss and staircase effect problems caused by the fixed parameters of total variation(TV ) regularization constraints in image spare reconstruction, this paper proposes an adaptive sparse image reconstruction algorithm by using second-order total generalized variation( TGV) model as the regularization constraints. The second-order TGV model is applied to balance the first and second derivative in images,and it can automatically modify the weights on the basis of each iteration solution and tensor func-tion to achieve image sparse reconstruction. Simulation results show that compared with the TV model and fixed TGV model,this algorithm can maintain both image detail information and image outline, as well as im-proving peak signal-to-noise ratio(PSNR) and structure similarity( SSIM) of the reconstructed image.
出处
《电讯技术》
北大核心
2017年第9期981-985,共5页
Telecommunication Engineering
关键词
图像处理
稀疏重建
压缩感知
广义全变分
自适应正则约束
分裂Bregman算法
image processing
sparse reconstruction
compressed sensing
total generalized variation
adap-tive regularization term
split Bregman algorith