Automatic detection and removal of baseline wanders is most important for accurate measurement of clinical features of local waves including, P-wave, QRS complex, T-wave and U-wave of the ECG signals. In this paper, we investigate on the use of variational mode decomposition (VMD) technique for removal of baseline wander in ECG signals. The proposed method consists of three stages: VMD based ECG signal decomposition, baseline wander extraction using frequency criterion, and base-line wander subtraction. The method is tested and evaluated using a wide variety of ECG signals including different kinds of PQRST morphologies corrupted with different patterns of baseline wanders. Performance evaluation results show that the VMD based method can adequately remove the baseline wanders without significantly distorting local components (P-wave, QRS complex, ST-segment, T-wave) of the ECG signal. The noise reduction capability of the VMD-based method is compared with the existing methods using the wavelet transform and Fourier transform. The subjective and objective quality assessment results show that the VMD based method outperforms the existing methods in terms achieving better objective quality metric values for various noisy ECG signals. © 2016 IEEE.