摘要
为从模糊图像中恢复出更多细节和纹理信息,提出一种基于结合全变差(TV)和分数阶全变差(FOTV)模型的数字图像去模糊方法。用全局梯度提取法将模糊图像分解成平滑区域、凸边和纹理3部分,用全变差模型约束平滑区域和凸边,用分数阶全变差模型约束细节部分,建立去模糊的凸优化模型,用变量分裂和交替方向法快速求解该模型。实验结果验证了该模型和求解算法的有效性和快速性,给出了每组实验的PSNR和SSIM值。
To recover more details and textures from blurred image,an image deblurring method based on combining total varia-tion and fractional-order variation models was presented.The blurred image was decomposed into constant regions,salient edges and details using a global gradient extraction scheme (GGES).The total variation model was applied to the constant regions and salient edges,and the fractional-order total variation model was applied to the details.A convex optimization model of image de-blurring was established.A variable split and alternative direction method was employed to solve the optimization model quickly. The results of experiments demonstrate the validity and efficiency of the proposed model and the algorithm,and the PSNR and SSIM of each group experiment are provided simultaneously.
出处
《计算机工程与设计》
北大核心
2016年第7期1857-1861,1866,共6页
Computer Engineering and Design
关键词
去模糊
全变差
分数阶全变差
凸优化
交替方向法
image deblurring
total variation
fractional-order total variation
convex optimization
alternative direction method