Fusion of infrared and visible images using empirical mode decomposition and spatial opponent processing

Infrared (IR) cameras capture thermal radiations emitted by objects in a scene while Electro-Optical (EO) cameras picture reflected colors from a scene. These sources of different modalities produce complementary data of the same panorama. In aeronautics, medical imaging and rescue operations, these sensors are often utilized simultaneously. The objective of this work is to integrate the content of both streams in order to deliver a unique image presenting more visual cues than any of the originals taken separately. To reach that goal, the synthesis of a robust algorithm is required. The Empirical Mode Decomposition (EMD) decomposes image signals into Intrinsic Mode Functions (IMFs). In this paper, we proposed a novel approach that integrates the IR and EO IMFs using principles of neural science for multi-spectral fusion. We show how to integrate IR and EO information utilizing spatial opponent processing. We report the performance on images from Octec ltd and compare our result to their Wavelet-based fusion output. Mutual Information is the metric employed for fusion quality assessment in this work. Positive results have been obtained for tests conducted on IR and EO images containing complementary information.