基于相对清晰度的多焦距图像融合算法

Multi-focus Image Fusion Algorithm based on Relative Activity Level

针对现有图像域融合算法难以合理度量区域清晰度,导致融合结果出现"振铃"现象的问题,提出一种基于图像区域相对清晰度的多焦距图像融合算法。该算法定义了图像区域的相对清晰度判定指标,并根据多焦距图像像素邻域的相对清晰度来逐个选取聚焦清晰的像素点从而构成融合图像。实验表明,所提算法能取得优于其它算法的融合效果;特别是,它能较好保留源图像包含的细节信息,从而显著减弱融合结果中的"振铃"现象。

Most existed region-based multi-focus image fusion methods adopt unreasonable region clarity （or focus） measures, which bring notable ringing phenomena to their results. To overcome this problem, a novel fusion algorithm is proposed, which is based on a new image region focus measure, Relative Activity Level （RAL）, presented in this paper. The algorithm chooses most focused pixels, in terms of the RAL of the pixels＇ neighborhood, from source images to form the final fusion results. Experiments prove the proposed algorithm to be superior to other methods. Especially, it can preserve details in source images better, thus reduce ringing phenomena in fused images remarkably.