Optimal image fusion using the Rayleigh quotient

The advancement in sensor technology has led to several sophisticated imaging sensors, which can be used to study the properties of an object of interest. However, in practice a single sensor is not enough to capture all the characteristics of the object. Therefore, a suite of sensors having complementary properties is often used to inspect the object. This necessitates the use of a fusion algorithm to integrate information from different sensors into a single entity for interpretation. In this paper a computationally efficient optimization based approach has been proposed for pixel level fusion. The proposed approach models the fusion process as a weighted average of input data. The weights of the fusion process are estimated by maximizing the average power of the corresponding fused pixel. It is shown that this optimization problem is equivalent to maximizing the Rayleigh quotient, and that the proposed approach maximizes the signal-to-noise ratio of the fused pixel. Results on a variety of databases clearly indicate the feasibility of the proposed approach.