Visualization of multispectral video with moving background based on background extraction and fusion

Detecting a target in a visible video is a challenge at times, as RGB sensors do capture color and correct relative luminance, but are under exposed and noisy. In the infrared band, visibility of the same target is better but they lack color and relative luminance. Frame by frame fusion of these two videos can be a solution but computationally expensive and thus impractical. For the video captured by fixed camera, the background can be extracted easily and fused with the moving target to save the computation. But if the camera is moving, then the extraction of the moving background is a big challenge. In this paper, we present an efficient multi spectral video visualization method using estimation of the background in motion. The analysis shows that the proposed method is superior compared to the existing methods quantitatively as well as visually. More importantly it is devised to use in real time to make it more practical.