1. 北京工业大学信号与信息处理研究室,北京,100124
2. Department of Electronic and Information Engineering,The Hong Kong Polytechnic University,Hong Kong
3. 北京工业大学信号与信息处理研究室北京,100124
4. Department of Electronic and Information EngineeringThe Hong Kong Polytechnic UniversityHong Kong
网络出版:2008-09-25,
纸质出版:2008
移动端阅览
李晓光, 沈兰荪, LAM Kin-Man, 等. 一种基于GVF各向异性扩散模型的图像放大算法[J]. 电子学报, 2008,36(9):1755-1758.
LI Xiao-guang, SHEN Lan-sun, LAM Kin-man, et al. An Image Magnification Method with GVF-Based Anisotropic Diffusion Model[J]. Acta Electronica Sinica, 2008, 36(9): 1755-1758.
提出了一种基于梯度向量流各向异性扩散模型的图像放大方法.首先低分辨率图像插值放大作为高分辨率图像的初始估计
然后利用基于GVF的平均曲率扩散模型和高斯移动平均低分辨率模型约束进行迭代复原.GVF是一种有旋场
作为外力场用来描述图像的边缘特征
能够将初始图像中斜向边缘锯齿效应表示为流线型.采用GVF外力场约束平均曲率扩散过程
能够有效去除边缘锯齿现象并保持纹理结构.高斯移动平均模型提供了图像数据保真度约束
使结果更接近理想图像.实验结果表明
本文算法能够有效提高放大图像的主观视觉质量和客观PSNR.
An image magnification method with GVF-based anisotropic diffusion model is proposed.An image is magnified by bilinear interpolation at first.Then
an iterative restoration with a GVF based mean curvature flow diffusion and a Gaussian moving average LR constraint is applied to the magnified image.Since GVF is a rotational field
as an external force field to descript the edges of an image
the vector flow will become streamline near the jagged edges.Therefore
the GVF based anisotropic diffusion will be helpful to remove the jagged effects as well as keep the texture structures.Meanwhile
the Gaussian moving average LR model provides a data fidelity constraint which makes the results more close to the ideal HR images.Experiments results show that the proposed method can improve the quality of magnified image in terms of both the objective and subjective.
0
浏览量
1146
下载量
1
CSCD
关联资源
相关文章
相关作者
相关机构
京公网安备11010802024621