Abstract: As some TH-1 images have the problem of big noise, a fusion method for TH-1 images based on FDCT was proposed. During the fusion processing, the sub-band of low frequency was processed by the weighted average algorithm; the sub-band of middle frequency, which has lower noise, was processed by the biggest region edge detection algorithm; the sub-band of the highest frequency, which has more intensive noise, was processed by the maximum substitution algorithm. The proposed method clearly outperforms traditional fusion methods in common image evaluation indicators and noise reduction evaluation indicators.

Key words: TH-1, fast discrete curvelet transforms, noise reduction, image fusion, image quality evaluation