For achieving a precise diagnosis of rolling bearings，post-processing is a commonly used practice to eliminate in-band noise remained by band-pass filtering in the context of resonance demodulation. Usually，the pre- and post-processing are optimized independently，and the optimization indices don’t take the periodical occurrence of impulses into account，which are the main problems with existing paradigms. Aimed at such deficiencies，a novel bearing diagnosis method is proposed by a consequent use of Morlet filter and maximum correlated kurtosis deconvolution（MCKD），where niching genetic algorithms（NGAs）determine the optimal parameters of such two steps in a synchronous fashion involving the center frequency and bandwidth of Morlet wavelet filter，as well as the length and period of MCKD filter. Meanwhile，correlated kurtosis（CK）serves as optimization index to depict the periodicity of impulses. Simulated signals and experimental data of rolling bearing faults demonstrate the effectiveness and advantages of the proposed methods.