In order to reduce the computational cost while maintaining the sufficient fusion quality, a novel notion of image fusion approach was explored combining fusion with data compression based on compressed sensing. First, the sensing data was compressed by random projection. Then, the sparse coefficients were obtained on compressed samples by sparse representation. Finally, the fusion coefficients were combined with the fusion impact factor and the fused image was reconstructed from the combined sparse coefficients. Experimental results validate its rationality and effectiveness, which can achieve comparable fusion quality on less compressed sensing data.