Abstract: Aimed at features of infrared and visible images and combined with V.Petrovic's fusion method,an improved approach based on wavelets on intervals is presented.Gradient filter of wavelets on intervals is first presented in matrix form.Then the gradient map of two original images is disposed with fusion/decomposition step and a multiresolution pyramid representation of original images is obtained.Averaging two low-pass components in the pyramid and taking inverse transform,the fusion image is obtained.Evaluated with subjective test and objective performance measurement,the algorithm is found to be practical,and the fusion result with biorthogonal wavelet on intervals is proved to be better than orthogonal wavelet on intervals,which accords with the fact that biorthogonal wavelet possess symmetry and is more suitable for image processing.