增强无人机态势感知的彩色图像融合方法

Color image fusion method for enhancing situation awareness of ICA

机载传感器侦察及地面数据融合处理是美军军用研究实验室规划的无人机态势感知技术发展的第四等级。独立成分分析(ICA)应用于图像处理是在分析人眼视觉系统特性基础上利用稀疏编码的一种新颖的变换域方法,具有多方向性、特征提取及边缘建模特性。色彩传递是目前融合图像自然感彩色化的最佳途径。以增强无人机态势感知为目的,结合两者研究突出波段特征的自然感彩色融合方法。根据场景建立训练图像库并提取独立波段特征信息构建ICA域的分析核和综合核,在ICA域按照区域能量融合规则生成灰度融合图像,根据视觉任务将源图像线性映射到色彩通道赋予灰度融合图像彩色信息,采用控向金字塔对源彩色融合图像和彩色参考图像的各通道进行多分辨率分解,将各通道的子图像进行均值、方差传递,最终获得类似彩色参考图像色彩的融合图像。人眼感知和客观评价表明:融合图像波段特征突出且增强了细节信息,色彩自然、舒适,进一步改善了机载平台的场景感知。

Airborne sensors reconnaissance and ground data fusion processing is the fourth grade of UAV situation awareness that planned by USA Office of the Secretary of Defense. Independent Component Analysis（ICA） that applied to the image processing field is a novel method of transform domain in the analysis of human visual system characteristics based on sparse coding theory, with multiple directions,excellent characteristic extraction and edge modeling feature. Color transfer is the best way to get natural sense color fusion image. The combination of studies highlighted the band characteristics of the natural sense color fusion method so as to enhance UAV situation awareness. Training image database was established according to the scene and the independent band feature information was extracted to construct ICA domain analysis kernel and synthesis kernel. In the ICA domain, the gray fusion image was generated applying area energy fusion rules, the gray fusion image color information was given using source image linear projection to the color channel. The various channels of source color fusion image and color reference image were multi-resolution decomposed using steerable pyramid, each channel transfer mean and variance were independently completed. Finally a similar color fusion image was obtained. Eye perception and objective evaluation show that outstanding band features and natural color enhance detail information to further improve the airborne platforms scene perception.