Real-time Multi-modal Image Fusion for Flight Vision System

In the field of intelligent avionics, Flight Vision System is a kind of advanced cockpit human-machine interface including Enhanced Vision System, Enhanced Flight Vision System and Combined Vision System, etc. The display output of Flight Vision System selectively integrates long/short wave, visible image sensor, Lidar/Radar, digital map images. With the help of multimodal sensor images which have the visibility see-through capacity in different visual condition, pilots achieve equivalent visual operations. However, in airborne computing platform, due to limited computing resources, only simple methods such as α-blending are used to obtain real-time performance. Usually the fusion result is often not satisfactory. In this paper, we improved the image layer separation based fusion method. On one hand, we employed the down sample version of fast Guided Filter as core image filter, on the other hand, we modified the main sections of original fusion algorithm in parallel fashion to improve the computing efficiency. Experimental results show that the proposed method keeps the good fusion appearance and reaches real-time performance with more than 30 fps, which is quite satisfactory for the application requirement in Flight Vision System.