摘要
经当前方法复原后图像与实际图像之间的吻合度较低,且复原效率低。为此提出基于微分方程的退化图像盲复原数学模型构建方法。通过贝叶斯萎缩方法对阈值选择进行优化,对退化图像做二维离散小波变换,并采用高频系数进行离散Curvelet变换。在此基础上对阈值进行估计,获得最优阈值,实现退化图像的去噪。根据去噪结果在高阶微分方程中引入稀疏测度、卷积滤波器和引导图像特征,通过范数约束获得强边引导图像,在潜在图像中应用稀疏函数和卷积滤波器获得梯度下降方向,得到强边图像,通过强边图像对模糊核进行估计,实现退化图像的盲复原。仿真结果表明,所提方法的复原效率高、复原程度高,说明该方法复原效果较好,应用价值较高。
At present, the consistency between the restored image and the actual image is low, and the restoration efficiency is low. Therefore, a mathematical model for blind restoration of degraded images based on differential equation was constructed. First of all, Bayesian shrinkage method was used to optimize the threshold selection, and the degraded images were transformed by two-dimensional discrete wavelet. After that, the high-frequency coefficient was adopted for discrete Curvelet transform. On this basis, the threshold value was estimated to obtain the optimal threshold value and thus to remove the noise from the degraded image. According to the denoising results, the sparse measure, convolution filter and guidance image features were introduced into the higher-order differential equation, and then the guidance image on strong side was obtained by norm constraint. In the potential image, sparse function and convolution filter were used to obtain the gradient descent direction and the image on strong side. The fuzzy kernel was estimated by the image on strong side. Finally, the blind restoration of degraded image was achieved. Simulation results show that the proposed method has high restoration efficiency, high restoration degree and high application value.
作者
姚海燕
YAO Hai-yan(Dongchang College,Liaocheng University,Shandong Liaocheng 252000,China)
出处
《计算机仿真》
北大核心
2021年第5期185-188,431,共5页
Computer Simulation
关键词
微分方程
退化图像
盲复原
Differential equation
Degraded image
Blind restoration
