A general and flexible approach for constructing two-dimensional multi-channel morphological wavelets was presented,which was based on templates and lifting scheme.The gradient templates were applied to extract detail signals in predicting lifting step,and all kinds of morphological operators (max,min,median and so on) were used as update operators to update the low-pass signals.In order to guarantee nonlinear perfect reconstruction,the necessary condition of choosing templates was discussed.The wavelets constructed on this new framework were nonlinear and suitable for image processing applications.Finally,some examples were given to reveal the advantages and applicability of this new method.