基于图像块能量梯度的多聚焦图像融合

Multi-focus Image Fusion Method Based on Energy Gradients of Image Blocks

摘要:

针对多聚焦图像,提出一种基于图像分块的融合方法。将源图像分为大小相同数量相等的子块,采用能量梯度算子作为对焦评价函数,计算各个图像子块能量梯度匹配度,设置匹配度阈值分离出源图像中的清晰区域。源图像中的清晰区域直接作为融合图像相应的区域,其它区域的处理中,构造与相应子块能量梯度大小相关的图像序列,以及像素点到各个子块中心距离相关的融合函数,然后用融合函数对图像序列融合。实验结果表明该方法有效性和合理性。

Abstract:

A new effective fusion method based on image blocks was proposed for multi-focus images. Firstly, two source images were divided into equal number of blocks with the same size. Then each block of two source images was inspected to be determined whether it is a clear block or not by analyzing the match degree of the energy gradient of the corresponding images blocks of source images, in which the energy gradient operator was chosen as the focus measure and a threshold value was set for the match degree of energy gradients. Then the clear domains of source images were isolated while the other regions were deemed fuzzy. All pixels in clear domains would be directly used as the corresponding ones of the fused image instead of any kind of processing in order to retain the in-focus regions as much as possible. However, for fusing the fuzzy regions of source images, an image sequence was constituted which depends on which source image has a higher average gradient for each image block, in which the image sequence contains a blending function was constituted based on the energy gradients of the image blocks that was relative to the distance of this pixel to the center of each image block. Then the image sequence and blending function were used to construct the fused image domains, in which the fused pixel was actually a weighted sum of the corresponding pixels in the images sequence. The experimental results shows that the method proposed is the reliability and rationality of image fusion results.