Multisensor colour image fusion for night vision

This paper presents two novel algorithms for colour fusion of multisensor night vision images that provide a natural encoding of thermal data in visible range imagery. In the first we use a simple method of direct encoding of RGB data with multisensor inputs while in the second we employ Hue-Saturation-Value colour space decomposition to treat intensity information in a natural manner but also encode unpopulated colour channels with non-visible range data. Conventional, robust multiresolution grayscale fusion of multisensor inputs is used to provide fused image monochrome intensity. Meanwhile relatively simple colour channel encoding transfers significant information from the thermal range image into relevant fused colour channels. Statistical normalisation avoids large scale false colouration that causes unnatural appearance of most pseudo-colour fused images. We demonstrate the advantage of our algorithm compared to state-of-the-art in false colour fusion on a number of well known multisensor sequences. We also analyse its implementation complexity and show that natural looking fused colour images can be obtained with limited computational resources.