基于各向异性热扩散方程的多聚焦图像融合算法

Multi-focus Image Fusion Algorithm Based on the Anisotropic Thermal Diffusion Equation

本文根据多聚焦图像的光学成像原理,利用各向异性热扩散方程来估计图像深度信息,并讨论一种基于深度信息提取的多聚焦图像的融合算法.算法首先对两幅多聚焦图像进行光学成像过程的模拟,分别建立正向区域的热扩散方程,通过对热扩散方程的迭代求解估计出图像场景的深度信息.然后对深度信息进行自适应的区域划分,分别得到每幅图像的清晰区域,模糊区域和中间过渡区域.最后通过提取清晰区域的像素点,融合过渡区域的像素点实现多聚焦图像的融合.理论推导和实验验证表明,本方法能够克服常用算法中出现的块效应和人工痕迹,是一种有效的图像融合算法.

A method for multi-focus image fusion based on depth information extraction is discussed. Under the optical imaging principle of multi-focus image,the anisotropic thermal diffusion equation is used to estimate the image depth information. Firstly,the optical imaging of two multi-focus images is simulated and the positive thermal diffusion equation is respectively established. After iteration of thermal diffusion equation,the image scene depth information is estimated. Depth information is adaptively divided,and images are divided into three regions： clear region,fuzzy region and intermediate transition region. Finally,the result of the proposed algorithm is achieved as soon as the pixels of clear region are extracted and pixels of transition region are fused. Theoretical analysis and experiment results showthat the proposed algorithm is an effective method of image fusion that can overcome the block effect and artifacts in general algorithms.