PET/CT medical image fusion has important clinical significance. As the multiwavelet transform has several particular advantages in comparison with scalar wavelets on image processing, this paper proposes a medical image fusion algorithm based on multiwavelet transform after in-depth study of wavelet theory. The algorithm achieves PET/CT fusion with wavelet coefficients fusion method. Experimental results show that fusion image combines information of the source images, adds more details and texture information, and achieves a good fusion result. Based on the proposed algorithm, we can obtain the best result when using gradient fusion in the low-frequency part and classification fusion in the high-frequency part.