This paper proposes a novel continuous wavelet transform (CWT) based approach to holistically estimate the dominant oscillation using measurement data from multiple channels. CWT has been demonstrated to be effective in estimating power system oscillation modes. Using singular value decomposition (SVD) technique, the original huge phasor measurement unit (PMU) datasets are compressed to finite useful measurement data which contain critical dominant oscillation information. Further, CWT is performed on the constructed measurement signals to form wavelet coefficient matrix (WCM) at the same dilation. Then, SVD is employed to decompose the WCMs to obtain the maximum singular value and its right eigenvector. A singular value vector with the entire dilation is constructed through the maximum singular values. The right eigenvector corresponding to the maximum singular value in the singular-value vector is adopted as the input of CWT to estimate the dominant modes. Finally, the proposed approach is evaluated using the simulation data from China Southern Power Grid (CSG) as well as the actual field-measurement data retrieved from the PMUs of CSG. The simulation results demonstrate that the proposed approach performs well to holistically estimate the dominant oscillation modes in bulk power systems.