Abstract: The existing image fusion methods usually set down the fusion rules before fusion processes.However,the rules which determine the attributes of fusion results could not be adjusted according to different future applications.In order to overcome this limitation,a framework for multi-focus image fusion is proposed based on data assimilation and genetic annealing algorithm.Under this framework,the wavelet transition method is regarded as the model operator,while the principal component analysis method is considered as the observation operator.The weights of different attributes are determined according to their different importance to the post-processing.The weighted sum of the evaluation indices is used to construct the object function.Finally,the object function is optimized by using genetic annealing to obtain the proper image.The experiments validate the effectiveness of the framework.