Hierarchical Optimization for MultiFundus Image Fusion
1 College of Computer and Information Fujian Agriculture and Forestry University，Fuzhou 350002, China
2 College of Physics and Information Engineering Fuzhou University，Fuzhou 350002, China
Abstract：In order to remove the seams and ensure information details are not missed in the image fusion, the multiband blending image fusion method for multifundus image was proposed in this paper based on hierarchical optimization fusion. The mask image is achieved by multithreshold segmentation
and Euclidean distance transform. And the improved weighted coefficients are designed based on the distance value in the overlap region and the layers of Gaussian image sum of Laplacian energy for the mask image. The combined objective evaluation was proposed that it is based on information entropy, spatial frequency and definition. The registration error and overlap rate are used to realize hierarchical optimization multiband image fusion method for grouping fusion. And the image alone is put into next level directly. 75 group image sequences are applied involving 4 898 set of images (normal fundus image 2 952 pairs and disease fundus image 1 946 pairs) for testing and validation. The results show that the proposed method is effective that removes the seam and obtains the RMSE value of about (01±005) pixels in overlap region. The proposed method can make equilibrium between objective evaluation and subjective visual effect.
［1］James AP, Dasarathy BV. Medical image fusion: a survey of the state of the art ［J］. Information Fusion, 2014,19:4-19.
［2］Xu Zhiping. Medical image fusion using multilevel local extrema ［J］.Information Fusion,2013,9(2014):38-48.
［3］张泾周,李婷,吴疆.医学图像的小波变换融合算法研究 ［J］.中国生物医学工程学报,2008,27(4):521-525.
［4］Cattin P, Bay H, Gool VL, et al．Retina mosaicing using local features ［C］ // Larsen R, Nielsen M, Sporring J, eds. Medical Image Computing and ComputerAssisted Intervention, Copenhagen: SpringerVerlag,2006, 〖STHZ〗4191〖STBZ〗:185-192．
［5］Daftari IK, Mishra KK, O’Brien JM, et al．Fundus image fusion in eyeplay software: an evaluation of a novel technique for ocular melanoma radiation treatment planning ［J］. Medical Physics,2010,
［6］Poletti E, Benedetti G, Ruggeri A. Superimage mosaic of infant retinal fundus: selection and registration of the bestquality frames from videos ［C］// Sergio Cerutti, Riccardo Barbieri, eds. Conference Proceedings IEEE EMBS. Osaka:IEEE,2013:5883-5886.
［7］魏丽芳,潘林,郑炳锟,等. 优化的多幅眼底图像拼接方法 ［J］.中国图象图形学报,2011,16(12):2175-2182.
［8］Choe TE, Cohen I, Lee M, et al. Optimal global mosaic generation from retinal images ［C］ //Rajapakse, eds. The 18th International Conference on Pattern Recognition, HongKong: IEEE Computer Society, 2006：6812-6841．
［9］Szeliski R. Computer Vision: Algorithms and Applications ［M］.New York: Springer,2010:27-59.
［10］Bailey DG．An efficient euclidean distance transform ［C］ // Reinhard K, Jovisa Z, eds. International Workshop on Combinatorial Image Analysis. Auckland: SpringerVerlag,2004:94-408．
［11］Wang Jun, Tan Ying. Efficient euclidean distance transform algorithm of binary images in arbitrary dimensions ［J］. Pattern Recognition,2013,46(1):230-242.
［12］屈小波,闫敬文,杨贵德．改进拉普拉斯能量和的尖锐频率局部化Contourlet域多聚焦图像融合方法 ［J］.光学精密工程, 2009，(5)：1023-1033．
［13］Ishita D, Bhabatosh C. Multifocus image fusion using a morphologybased focus measure in a quadtree structure ［J］. Information Fusion, 2013,14(2):136-146.
［14］Padma G, Vinod K. CT and MR image fusion scheme in nonsubsampled contourlet transform domain ［J］. Journal of Digital Imaging,2014,27(3):407-418.
［15］Wang Zhaobin, Ma Yide, Gu J. Multifocus image fusion using PCNN ［J］. Pattern Recognition, 2010, 43(6):2003-2016．
［16］阿都建华,王邦平,王珂,等.基于剪切波变换的医学图像融合算法［J］.中国生物医学工程学报,2013,(3):284-291.
［17］Piccinini F, Bevilacqua A, Lucarelli E. Automated image mosaics by nonautomated light microscopes: the MicroMos software tool ［J］. J Microsc, 2013, 252(3):226-250.
