一种基于深度神经网络模型的多聚焦图像融合方法

doi:10.6040/j.issn.1672-3961.2.2015.106

摘要/Abstract

摘要： 基于多聚焦图像融合中存在的低频信息易产生缺失的现象进行分析,提出一种基于深度神经网络模型的低频子带融合策略,并结合小波核滤波器及针对高频子带的融合策略,给出多聚焦图像融合方法。该方法利用自动编码器提取低频子带特征,利用网络隐层中的权值信息选择低频子带分量。采用3组聚焦不同的自然图像及1组医学图像进行算法测试,并与传统的低频子带融合策略进行对比,同时比较基于轮廓波变换的多聚焦图像融合方法、基于非下采样轮廓波变换的多聚焦图像融合方法。试验结果表明:其中一组图像采用深度神经网络模型的策略所得到的融合结果的边缘融合指标值能够达到0.802 7,优于其余比较方法的0.761 4、0.722 7和0.716 4,从而证实基于深度神经网络模型的融合策略的有效性。

关键词: 图像融合, 多聚焦图像, 小波核滤波器, 自动编码器, 深度神经网络

Abstract: There existed low-frequency information distortion phenomenon in fusing multi-focus images. Aimed to solve the problem, a new fusion strategy based on deep neural network model was proposed for fusing low-frequency subbands. Combined with Wavelet Kernel Filter and traditional fusion strategy for high-frequency subbands, a new fusion method for fusing multi-focus images was given. The method extracted efficient features by using AutoEncoder model. The experimental results showed that proposed method could obtain better images. The edge fusion qualify value of the proposed fusion result was 0.802 7, compared with traditional fusion strategy, contourlet-based multi-focus method and non-sampled contourlet-based multi-focus method, 0.761 4, 0.722 7, and 0.716 4, which could provide an effective method for fusing multi-focus images.

Key words: autoencoder, deep neural network, image fusion, multi-focus image, wavelet kernel filter

中图分类号:

TP183

刘帆,陈泽华,柴晶. 一种基于深度神经网络模型的多聚焦图像融合方法[J]. 山东大学学报(工学版), 2016, 46(3): 7-13.

LIU Fan, CHEN Zehua, CHAI Jing. A new multi-focus image fusion method based on deep neural network model[J]. JOURNAL OF SHANDONG UNIVERSITY (ENGINEERING SCIENCE), 2016, 46(3): 7-13.

参考文献

[1] POHL C, VAN GENDEREN J L. Multisensor image fusion in remote sensing: concepts, methods and applications[J]. International Journal of Remote Sensing, 1998, 19(5):823-854.
[2] BURT P J, ADELSON E H. The Laplacian pyramid as a compact image code[J]. IEEE Transactions on Communication, 1983, 31(4): 532-540.
[3] ZAVORIN I, LE MOIGNE J. Use of multiresolution wavelet feature pyramids for automatic registration of multisensor imagery[J]. IEEE Transactions on Image Processing, 2005, 14(6): 770-782.
[4] REDONDO R, ŠROUBEK F, FISCHER S, et al. Multifocus image fusion using the log-Gabor transform and a multisize windows technique[J]. Information Fusion, 2009, 10(2):163-171.
[5] BORWONWATANADELOK P, RATTANAPITAK W, UDOMHUNSAKUL S. Multi-focus image fusion based on stationary wavelet transform and extended spatial frequency measurement[C] //IEEE International Conference of Electronic Computer Technology. Macau, China: IEEE Press, 2009: 77-81.
[6] DO M N, VETTERLI M. The Contourlet transform: an efficient directional multiresolution image representation[J]. IEEE Transactions on Image Processing, 2005, 14(12): 2091-2106.
[7] YANG S Y, WANG M, JIAO L C. Image fusion based on a new contourlet packet[J]. Information Fusion, 2010, 11(2): 78-84.
[8] DA CUNHA A L, ZHOU J P, DO M N. The nonsubsampled contourlet transform: theory, design, and applications[J]. IEEE Transactions on Image Processing, 2006, 15(10):3089-3101.
[9] ZHENG S, SHI W Z, LIU J, et al. Multisource image fusion method using support value transform[J]. IEEE Transactions on Image Processing, 2007, 16(7):1831-1839.
[10] YANG B, LI S T. Multifocus image fusion and restoration with sparse representation[J]. IEEE Transactions on Instrumentation and Measurement, 2010, 59(4): 884-892.
[11] PETROVIC V. Subjective tests for image fusion evaluation and objective metric validation[J]. Information Fusion, 2007, 8(2): 208-216.
[12] BENGIO Y. Learning deep architectures for AI[J]. Journal Foundations and Trends^® in Machine Learning, 2009, 2(1): 1-127.
[13] HINTON G E, OSINDERO S, TEH Y W. A fast learning algorithm for deep belief nets[J]. Neural computation, 2006, 18(7):1527-1554.
[14] BENGIO Y, LAMBLIN P, POPOVICI D, et al. Greedy layer-wise training of deep networks[C] //Advances in Neural Information Processing Systems 19(NIPS'06). Vancouver, Canada: MIT Press, 2007: 153-160.
[15] HINDON G E, SALAKHUTDINOV R R. Reducing the dimensionality of data with neural networks[J]. Science, 2006, 313(5786): 504-507.
[16] COATES A, NG A Y, LEE H. An analysis of single-layer networks in unsupervised feature learning[M]. Brookline: Microtome Publishing, 2011: 215-223.
[17] ZHENG W L, ZHU J Y, PENG Y, et al. EEG-Based emotion classification using deep belief networks[C] //IEEE International Conference on Multimedia and Expo. Chengdu, China: IEEE press, 2014: 1-6.
[18] TANG J X, DENG C W, HUANG G B, et al. Compressed-domain ship detection on spaceborne optical image using deep neural network and extreme learning machine[J]. IEEE Transactions on Geoscience and Remote Sensing, 2015, 53(3): 1174-1185.
[19] HUANG P H, HUANG Y, WANG W, et al. Deep embedding network for clustering[C] //22nd International Conference on Pattern Recognition. Stockholm, Sweden: IEEE press, 2014: 1532-1537.
[20] LIU F, JIAO L C, YANG S Y. SAR image despeckling based on wavelet kernel transform and gaussian scale mixture model[C] //2nd Asian-Pacific Conference on Synthetic Aperture Radar. Xi'an, China: IEEE Press, 2009: 1088-1091.
[21] HUANG X S, CHEN Z. A wavelet-based image fusion algorithm[C] //IEEE Region 10 Conference on Computers, Communications, Control and Power Engineering. Beijing, China: IEEE Press, 2002: 602-605.
[22] 王海晖, 彭嘉雄, 吴巍, 基于小波包变换的遥感图像融合[J]. 中国图像图形学报, 2002, 7(9): 922-937. WANG Haihui, PENG Jiaxiong, WU Wei. Remote sensing image fusion using wavelet packet transform[J]. Journal of Image and Graphics, 2002, 7(9):922-937.
[23] 刘贵喜, 杨万海. 基于小波分解的图像融合方法及性能评价[J]. 自动化学报, 2002, 28(6):927-934. LIU Guixi, YANG Wanhai. A Wavelet decomposition-based image fusion scheme and its Performance Evaluation[J]. ACTA Automatica Sinica, 2002, 28(6):927-934.
[24] WANG Z, BOVIK A C. A universal image quality index[J]. IEEE Signal Processing Letters, 2002, 9(3):81-84.
[25] PIELLA G, HEIJMANS H. A new quality metric for image fusion[C] //International Conference on Image Processing. Barcelona, Spain: IEEE Press, 2003:III-173-6 vol.2.

多维度评价

Viewed

Full text

Abstract

Cited

Shared

Discussed