基于显著性映射与兴趣点凸壳的图像融合算法

Image fusion algorithm based on saliency mapping and interest points convex hull

提出了一种基于显著性特征的可见光与红外图像融合算法来改善目标的融合质量.引入显著检测器对红外图像进行处理,生成显著映射;进一步分析红外图像并检测兴趣点,提取图像中的显著兴趣点;通过计算显著兴趣点的凸壳确定显著区域;利用显著兴趣点凸壳对初始显著映射进行优化,使目标定位更加精确。根据区域映射获取可见光图像的背景区域;根据不同的融合准则对目标、背景区域进行融合,获得最终的融合图像.结果表明与当前可见光图像融合技术相比,所提算法在标准差、联合熵与边缘信息因子等指标方面具有优势,其融合图像的细节纹理更清晰。

A visible and infrared image fusion algorithm based on saliency features is proposed to improve fusion quality of objects. The saliency detector is introduced to process the infrared image for generating the saliency map. The salient interest points are extracted in the image by further analyzing the IR image and detecting interest points. The saliency areas are identified by calculating the convex hull of salient interest points, and the target positioning becomes more accuracy by using salient interest points to optimize the initial salient mapping. The background area of visible light image is obtained according to the region mapping, and the final fusion image is obtained by using different fusion rules to fuse the target and background regions. Results show that this algorithm has advantages with clearer detail texture of fusion image in standard deviation, joint entropy and edge information factor indexes compared with the common visible image fusion technique.