真彩色传递双波段图像融合

True color transfer for dual band image fusion

针对基于颜色传递的双波段(中波红外、可见光)图像融合时,由于参考图像选择不当,造成的彩色融合图像对比度差、图像边缘细节下降等问题,提出了基于拉普拉斯金字塔分解及对比度可调的色彩传递图像融合算法。对待融合的两幅图像进行拉普拉斯金字塔三级分解,每一级按照不同的融合规则进行融合,以重构后的灰度融合图像作为融合结构的亮度分量Y,将可见光与红外图像的差值图像作为U分量,以红外图像为V分量。在颜色传递公式中,加入可调的比例系数,通过适当调节比例系数可调节融合图像的对比度。实验结果表明,获得的融合图像较好地保留了可见光图像的细节,既突出了红外目标同时又具有参考图像的颜色特征。图像评价结果表明,融合图像的均值、方差、熵等评价指标均得到较大提高。

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.