A Wavelet-based Fusion Algorithm of Visual and IR Images in Scene Matching

摘要

A pixel level image fusion algorithm of merging the visual and infrared scene images is proposed to combine the complementary information of both kinds of images. This approach is based on wavelet transform, and uses two different fusion rules at the high frequency detail images and the coarsest approximation images. Both of the visual and IR images are first decomposed using discrete wavelet transform. Couples of high frequency detail subbands of wavelet coefficients from the source images are then locally merged according to the salience features. Meanwhile the corresponding gross approximations of source images at the coarsest resolution level are combined together according to the local matching measure between them. Reconstruction is then accomplished by inverse wavelet transform from the combined coefficients, and the final fused image is obtained.