一种基于网格变形的图像放缩方法  

A Content-Aware Image Retargeting Method Based on Mesh Deformation

作　　者：崔普明[1] 刘秀平[1] 王小超[1] 李淑华[1] 

机构地区：[1]大连理工大学数学科学学院,大连116024

出　　处：《计算机辅助设计与图形学学报》2013年第6期788-793,共6页Journal of Computer-Aided Design & Computer Graphics

基　　金：国家自然科学基金重点项目(U0935004);国家自然科学基金(61173102)

摘　　要：基于内容的图像放缩的关键是如何保持图像中重要的内容,同时最小化整幅图像的视觉变形.文中将图像放缩问题转化成网格变形问题,提出一种改进的基于内容的图像放缩方法.首先在网格上建立保持整体形状的能量函数,结合约束条件,通过求解凸二次规划问题得到变形后的网格坐标,最终实现图像的放缩.实验结果表明,该方法不仅能够保持图像中重要的内容和特征结构,而且较好地保持了图像的整体性,使得图像视觉变形较小.The most challenging issue for content-aware image retargeting is how to preserve the important content and minimize the visual distortions on the image structures. In this paper, we present an improved method for content-aware image resizing by leveraging a global optimization process. We reformulate the problem of image retargeting to a mesh deformation task and build a global shape preserving energy function on the grid. By solving a convex quadratic program with constraints, the deformed mesh coordinates are obtained, yielding the final retargeting image. Experimental results demonstrate that this approach feature structures, but also preserves the global image not only preserves the important content and configuration with low visual distortion.

关 键 词：图像放缩 显著性度量 整体结构 网格变形 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]