Enhancing Image Fusion – Pseudo Colour Multi-Scale Approach

Abstract

The existing system approach concentrates on the colour fusion of both day and night imagery’s, but they are less naturalistic for images containing regions that differ significantly in the coloured content. Even though colour mapping methods matches the statistical properties of the both day and night images, the accuracy of the colouring is very much depended on how well target and source images matches. To achieve accuracy, of colouring and naturalistic fusion this method uses a multi-scale scheme for image fusion using EM algorithm and pseudo colour-transfer algorithm. The fusion algorithm accomplishes YUV space. The Y component is obtained from the result of IR and visible image fusion, by which the original images are fused in non sub sampled contour let domain by using EM algorithm. U and V components come by a ratio of local to global divergence of the IR and visible image. The experimental result demonstrate that a better fusion performance with compare to traditional colour image fusion methods, which meets the requirement of accuracy colouring and naturalistic colour fusion.

Keywords: EM - Expectation and Maximization algorithm. NSCT - Non sub sampled contour let transformation algorithm, SAGE – Space Alternating Generalized Expectation Maximization, PCA - Principal Component Analysis, FSD - Filter Subtract Decimate.