In the application, the single wavelet may directly carries on the filter using the decomposition and the heavy construction formula to the signal. But the multi- wavelets are carried on the decomposition and the heavy construction with the vector filter group to the signal. The filter object must be a vector signal satisfied the certain request. Therefore, before carrying the multi- wavelets decomposition, We must pretreat the primitive discrete signal through the predicting filter to obtain the initial vector. After that it can be transformed using the multi- wavelets. Similarly, The heavy construction data should be post-process that can obtain the need results. The different multi- wavelets have the different pretreatment and the post-processing method. This article take the GHM multi- wavelets as an example, uses it in the image compression after the front filter, and compares its result with the single wavelet, obtains a better result