Abstract: Taking into account the characteristics of the infrared and visual images, for insufficient information content of the fused images, a novel algorithm for infrared and visible light image based on object extraction and NSCT is proposed. And the local information entropy based fusion ruler is used to process the high-frequency coefficients of object. Four fusion methods wavelet transform, Laplace-pyramid transform, NSCT, lifting directionlet transform, are selected as benchmark methods. And the fused image is quantitative analysed by mean of three indicatrix：information entropy, standard deviation, correlation coefficient. Experimental results indicate that proposed method not only solves the original image fusion algorithm for the problem of insufficient information content, but also  extracts the source image characteristics more effectively and accurately, it achieves better fusion effect in the subjective and objective evaluation.

Key words: object extraction, infrared image, fusion, Non Subsampled Contourlet Transform（NSCT）, region growing

摘要： 针对目前红外图像和可见光图像融合中，融合图像信息量不足的问题，将目标提取和NSCT方法相结合，对其中的高频目标区域提出了基于局部信息熵的融合规则。将其与小波变换法、拉普拉斯法、NSCT法、提升方向波变换法作比较，并通过熵、标准差、相关系数等参数对融合后的图像进行定量分析。实验结果表明，该方法不但较好地提高了融合图像信息量，而且能够更加有效、准确地提取源图像中的特征，在主观视觉效果与客观评价指标上均取得了较好的融合效果。

关键词: 目标提取, 红外图像, 融合, 非下采样轮廓波变换（NSCT）, 区域生长