摘要
针对传统图像放大处理过程中基于线性插值方法通常导致边缘模糊问题,分析了Tikhonov模型、全变差模型和高阶偏微分模型在图像处理中的优缺点,提出了一种全变差和高阶偏微分模型自适应结合的图像放大模型及推导算法。该模型对图像非平滑区域采用全变差模型处理,而平滑区域则采用高阶偏微分模型处理,最终新插入的图像点象素值由该点邻域象素自适应地各向异性加权得到,在保持图像边缘锐度的同时有效克服了平滑区域的阶梯效应。4种模型的实验比较验证了本文算法的有效性。
An efficient scheme is presented to overcome the edge blurring problem in tradition image zooming method using linear interpolation. Based on analysis the shortages in Tikhonov, total variation and higher order partial difference models, a novel algorithm is proposed by combining the total variation model and high-order PDE ones. The combined technique is able to preserve edges and avoide the staircase effect in smooth regions. Experimental results show the advantage of our proposed model compared with others.
出处
《南京邮电大学学报（自然科学版）》
2008年第5期79-83,共5页
Journal of Nanjing University of Posts and Telecommunications：Natural Science Edition
基金
江苏省博士后科研资助基金(0801019C)
南京邮电大学引进人才科研启动基金(NY207022)资助项目
关键词
图像放大
非线性扩散
全变差
高阶偏微分
image zooming
nonlinear diffusion
total variation
high-order PDE
