基于区域一致性的多聚焦图像融合算法  

Multi-Focus Image Fusion Algorithm Based on Regional Consistency

作　　者：王建[1] 王国珲[2] 王倩丽[1] 

机构地区：[1]西安航空学院计算机工程系,西安710077 [2]西安工业大学光电工程学院,西安710032

出　　处：《兵工自动化》2013年第4期55-57,共3页Ordnance Industry Automation

基　　金：陕西省教育厅科研计划项目资助(11JK0996)

摘　　要：针对同一场景两幅严格配准的多聚焦图像融合问题,设计一种基于区域一致性的多聚集图像融合算法。该算法利用比率金字塔变换得到图像的多分辨序列,采用基于区域一致性的融合方法,在图像的相应层次序列的各级金字塔图像上进行融合以获取最终的融合图像。从目视效果来看,采用文中算法进行图像融合,基本上能够将两幅左右聚焦的图像融合成一幅清晰的单一聚焦图像。实验结果表明:该算法与同类其他算法相比较,相关系数提高了0.08%,偏差度降低了30.67%～31.05%,互信息提高了7.92%～7.97%。Focusing on the fusion problem of the multi-focus images from same scene, the multi-focus image fusion algorithm based on regional consistency is designed. By using multi-resolution image sequences obtained from the ratio pyramid transform, the source images were fused on each corresponding levels with the regional consistency method, then the fused image was achieved through inverse ratio pyramid transform. From the visual effect, the designed algorithm can fuse into a clear single focus image from left-focus image and right-focus image. Experimental results and comparisons with three traditional algorithms demonstrate that the designed algorithm is superior with 0.08% improvement of correlation coefficient, 30.67% to 31.05% reduction in the deviation degree and 7.92% to 7.97% enhancement in mutual information.

关 键 词：多聚焦图像 比率金字塔变换 图像融合 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]