True color transfer for dual band image fusion

Abstract

Because of the inappropriate choose of reference image,double band( midwave infrared and visible) color fusion image based on the color transfer method has low contrast,decreasing image edge details and the infrared target is not prominent. We present a new algorithm based on the Laplacian Pyramid,and the contrast could be adjusted anywhere. We use the Laplacian reconstructed image as the luminance component of the YUV image fusion constructure,the differential image of visible image and infrared image as the U component,and the infrared image as the V component. In the color transfer formula,the adjustable scaling factor is brought in to adjust the contrast of fusion image. Experiments show that by using the improved method,the color fusion image retains meaningful details. The infrared target is prominent and the fusion image has similar color distribution with the reference image. We evaluate the quality of fusion image by using objective methods,and the result shows that evaluation indexes such as the mean,variance,entropy,average gradient and spatial frequency are improved.