摘要
由散焦图像恢复三维景物的深度信息是一个不适定问题.提出一种新的基于整体变分的散焦图像深度恢复算法:首先将散焦图像深度恢复转化为带有整体变分正则化项的能量泛函极值问题,然后采用变分原理将其中的最小化问题转为偏微分方程的求解,最后通过方程迭代获得深度的最优解.该算法避免了解不适定问题的逆,恢复聚焦图像等问题.模拟图像和真实图像的实验结果表明该算法是有效的,与最小二乘法相比具有较小的均方根误差.
The depth recovery of objects from defocus images is an ill-posed problem. We propose a novel depth recovery algorithm based on total variation. The algorithm includes three main processes : converting the depth recovery problem into the energy functional minimum problem with a regularization term of total variation, changing the minimum problem into the solution of partial differential equations by the principle of variation and obtaining the depth information by iterative procedures. We need not solve an inverse ill-posed problem and recover a focus image. Using synthetic and real images, we show that the algorithm is quite effective, leading to lower root-mean-souared error in comparison with other method.
出处
《小型微型计算机系统》
CSCD
北大核心
2011年第3期543-546,共4页
Journal of Chinese Computer Systems
基金
国家自然科学基金项目(60603083
60872106)资助
安徽大学博士科研启动经费项目(02303319)资助
关键词
深度恢复
整体变分
偏微分方程
正则化
depth recovery
total variation
partial differential equation
regularization