刘琰煜, 周冬明, 聂仁灿, 侯瑞超, 丁斋生. 低秩表示和字典学习的红外与可见光图像融合算法[J]. 云南大学学报(自然科学版), 2019, 41(4): 689-698. doi: 10.7540/j.ynu.20180753
引用本文: 刘琰煜, 周冬明, 聂仁灿, 侯瑞超, 丁斋生. 低秩表示和字典学习的红外与可见光图像融合算法[J]. 云南大学学报(自然科学版), 2019, 41(4): 689-698. doi: 10.7540/j.ynu.20180753
LIU Yan-yu, ZHOU Dong-ming, NIE Ren-can, HOU Rui-chao, DING Zhai-sheng. Infrared and visible image fusion scheme using low rank representation and dictionary learning[J]. Journal of Yunnan University: Natural Sciences Edition, 2019, 41(4): 689-698. DOI: 10.7540/j.ynu.20180753
Citation: LIU Yan-yu, ZHOU Dong-ming, NIE Ren-can, HOU Rui-chao, DING Zhai-sheng. Infrared and visible image fusion scheme using low rank representation and dictionary learning[J]. Journal of Yunnan University: Natural Sciences Edition, 2019, 41(4): 689-698. DOI: 10.7540/j.ynu.20180753

低秩表示和字典学习的红外与可见光图像融合算法

Infrared and visible image fusion scheme using low rank representation and dictionary learning

  • 摘要: 针对目前红外与可见光融合算法在保留可见光图像中的背景信息时无法同时有效地提取红外图像信息,提出了一种基于低秩表示和字典学习的红外与可见光的图像融合算法. 首先,采用低秩表示对红外图像和可见光图像进行分解,分别获得源图像的低秩和稀疏成分,其中稀疏成分可以很好地表示源图像的边缘细节特征. 其次,用OMP算法的字典学习方法和稀疏系数的最大范数规则,而最大范数规则在对图像背景恢复的同时能够提取目标信息.再次,对分解得到的2个分量进行融合. 最后,利用融合稀疏系数和自适应字典重建融合图像. 实验结果表明,本融合算法可以突出红外对象信息,同时能够保留可见光图像中的背景信息,达到良好的视觉效果.

     

    Abstract: Current infrared and visible images fusion algorithms couldn't efficiently extract the object information in the infrared image while retaining the background information in visible image. To solve this problem, a new infrared and visible image fusion algorithm by low rank representation and dictionary learning was proposed to promote contrast and preserve edges for the source images. Firstly, low rank decomposition was performed on the input images to obtain their corresponding low rank and sparse components which could well represent the sparse feature of images. Secondly, the sparse representation using OMP algorithm with a trained dictionary was adapted to calculate the low rank coefficient and sparse coefficient, then by adding the common low rank sparse coefficient to the maximum norm of uncommon sparse coefficients could retain the background information of the source image effectively. Finally, we reconstructed the fused image from the fused sparse coefficients and adaptive dictionary. Experimental results demonstrated that this fusion algorithm could highlight the infrared objects and retained the images detailed information as well as retain the background information in visible image.

     

/

返回文章
返回