视觉显著性指导的红外与可见光图像融合算法

为了获取适合人眼观测的高质量红外与可见光融合图像,提出了一种基于视觉显著性指导的红外与可见光图像融合算法。首先,利用改进的流形排序法分别检测红外与可见光图像的视觉显著性区域;然后,采用非下采样轮廓波变换对红外和可见光图像进行多尺度、多方向分解,从而获取各自低频子带和高频子带,并将视觉显著性的检测结果用于指导分配低频子带的融合权重,即依据显著度大小赋予不同的权值,而高频子带的融合则依据局部标准差准则赋值;最后,通过非下采样轮廓波逆变换获得融合图像。实验结果表明:这种算法不仅可以保全可见光图像中的细节信息,而且能够精确地突显出红外目标信息,具有较好的视觉效果, 增强了红外与可见光复合前视系统的识别性能。

Fusion of infrared and visual images guided by visual saliency

To obtain a high quality fused image consistent with characteristics of human vision, a novel image fusion method for infrared and visual images guided by visual saliency is proposed. First of all, for the given infrared and visible images, the modified Manifold Ranking algorithm is utilized to extract their visual salient areas respectively. Then, source images are decomposed in different scales and directions by Non-subsampled Contourlet Transform to obtain low frequency information and high frequency information. And results of visual saliency detection are used to guide the fusion rule of low frequency subband coefficients. Besides, the high frequency subband coefficients are fused owing to the local standard deviation criterion. Finally, the fused image is obtained by performing inverse Non-subsampled Contourlet Transform. Experimental results demonstrate that the proposed algorithm can not only assure the final fused images with clear detail information, but also highlight the infrared objects accurately, which presents a good vision effect and effectively enhances recognition probability of infrared and visible compound systems.