To make up the shortcomings that data redundancy is large and the fusion result is not ideal when pyramid transform algorithm decomposes, a new algorithm of remote sensing image fusion optimized by pyramid transform algorithm is proposed. The proposed algorithm uses pyramid decomposition to build pyramid sequence. According to the relevant weight coefficient given by prior knowledge, the remote sensing image is reconstructed by iterating. Then through optimization selection of Baldwinian clonal selection algorithm, within the acceptable scope of iteration, the weight coefficient is adaptively modified and chose, and the suitable fusion parameter is sought and estimated to optimize fusion effects so as to avoid empirical selection of pyramid transform algorithm. In order to highlight the algorithmic merits, experiment applies pyramid transform optimization, pyramid transform optimization via genetic algorithm, and pyramid transform optimization via particle swarm optimization algorithm to make a comparison. Fusion quality is analyzed and evaluated from the two aspects of visual effect and mathematical statistics. Experimental result indicates that this arithmetic fusion result by the proposed method is consistent with human visual perception, and conducive to image analysis and information extraction.