多视觉特征和引导滤波的鲁棒多聚焦图像融合

发布时间：2018-08-28 15:14

【摘要】：为建立一个有效的活跃度测量模型来检测多聚焦图像的聚焦区域,针对多数融合方法效率不高和处理源图像未配准问题的不足,提出一种基于多视觉特征和引导滤波的快速鲁棒多聚焦图像融合方法.首先通过分别测量对比度显著性、清晰度和结构显著性这3个互补的视觉特征对源图像聚焦区域进行检测,获得初始的融合决策图;为了充分利用空间一致性并有效抑制融合结果中伪影的产生,利用形态学滤波和引导滤波对初始决策图进行优化,从而获得最终的融合权重图;最后根据优化的权重图对源图像进行加权融合,获得融合图像.实验结果表明,无论是主观视觉效果还是客观定量评价,该方法均优于一些主流的多聚焦图像融合方法.
[Abstract]:In order to establish an effective activity measurement model to detect the focus region of multi-focus image, the efficiency of most fusion methods is not high and the problem of unregistration of source image is insufficient. A fast robust multi-focus image fusion method based on multi-visual features and guided filtering is proposed. Firstly, three complementary visual features, namely, contrast saliency, clarity and structural significance, are measured to detect the focus region of the source image, and the initial fusion decision map is obtained. In order to make full use of spatial consistency and effectively suppress artifacts in fusion results, the initial decision map is optimized by morphological filtering and guided filtering, and the final fusion weight map is obtained. Finally, the source image is weighted and fused according to the optimized weight graph. The experimental results show that the proposed method is superior to some mainstream multi-focus image fusion methods both in subjective visual effect and objective quantitative evaluation.
【作者单位】：江西财经大学信息管理学院;江西省数字媒体重点实验室;江西财经大学软件与通信工程学院;
【基金】：国家自然科学基金(61662026,61462031,61262034) 江西省自然科学基金(20151BAB207033,20161ACB21015) 江西省高校科技落地计划项目(KJLD14031,GJJ150461)
【分类号】：TP391.41

【参考文献】