基于CV模型和NSCT的红外与可见光图像融合  

Image fusion based on CV model and NSCT for visual and infrared images

作　　者：毕家宾[1] 牛珍珍[1] 魏宗寿[1] 

机构地区：[1]兰州交通大学光电技术与智能控制教育部重点实验室,兰州730070

出　　处：《计算机工程与应用》2013年第19期180-185,共6页Computer Engineering and Applications

摘　　要：针对传统融合方法在光照不足、目标隐藏或目标和背景颜色接近时,容易出现目标信息丢失或减弱的现象,提出一种将基于CV模型的目标提取与NSCT相结合的方法。该方法使用动态轮廓线模型对红外目标进行搜索检测识别,将源图像序列分为目标和背景区域,利用非下采样Contourlet变换对输入图像进行多尺度、多方向稀疏分解,准确捕获图像中的高维奇异信息,并在目标和背景区域里分别采用不同的融合规则,将其与小波融合方法、拉普拉斯融合方法、NSCT方法作对比,并通过熵、平均梯度、空间频率、标准差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的目标探测性,而且融合结果中的目标比较清晰,亮度较高,目视效果较好,在主观视觉效果与客观评价指标上均取得了很好的融合效果。In order to solve the problem that the target information is easily lost or impaired when the target is hided or the color of target is close to the background or lack of light, a new method that combined NSCT and target extraction based on CV model is proposed. The method uses the dynamic contour model to search, test and identify the infrared target, then divides the source image sequence into target and background region, and uses non-subsampled Contourlet transform to sparse decomposition in multi-scale and multi-directions for getting high-dimensional singular information accurately. Different fusion rules are used in target and background region respectively, and this method is compared with wavelet fusion method, Laplace fusion method, and NSCT fusion method. Quantitative analysis is carried out for the fused image under parameters like entropy, average gradi-ent, spatial frequency and standard deviation. The results show that this method can not only make the detection of fusion target more easily, but also make the target look clear and brighter. A good fusion effect in subjective visual andobjective evaluation index is obtained.

关 键 词：CV模型 目标提取 NSCT方法 图像融合 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]