In physics class, the residual data before and after noise elimination contains more image information. In the equivalent wave domain, the K-SVD method is used to eliminate noise in the noise image of each frequency band and the residual of each frequency band. The main objective is to improve the image quality of noise. The method is combined with the denoised sub-spectrum to obtain the denoised sub-spectrum with the residual signal. An inverse isomorphic wave is used for noise reduction. The simulation results show that the noise reduction algorithm using K-SVD is better than the equivalent waveform and K-SVD mode. This method has achieved good results in practical application.