Multi-spectral imaging system can simultaneously obtain features of imag e and spectrum,and integrate them into a high quality image.Overall multi-spectral im age brightness is low,contrast is low,and the target and the background to distinguish are not obv ious.For dispersion- type multi-spectral imaging system,different wavelengths of light have differen t refractive indices,leading to every image of different bands focusing in different position s.So for all bands,collecting a group of images at one position will affect the multi-spectral′s focusing accu racy.The current digital image focusing algorithm based on optical theory sets the ideal focused image after a degradation process to form a focal and noisy image.Th is paper considers the coke degradation model,forms frequency domain field beginning,ch ooses Gaussian point spread function inversion model to do full digital autofocusing f or multi-spectral images,and multiplies them by a low pass filter to reduce noise from high freque ncy component.The results show that this algorithm can keep the energy of spectrum and contrast r atio,enhance details,and reduce noise at the same time,to make the details of original images more clear which are close to the focus images.Based on the above researches,this paper also uses hig h performance objective evaluation function,gray difference function,to evalute this filter′ s validity in subjective and objective aspects.