The very first step to process electrocardiogram (ECG) signal is to eliminate baseline wandering interference that is usually caused by electrode-skin impedance mismatch, motion artifacts due to a patient\'s body moment or respiratory breathing. A new method is thus suggested to remove baseline wandering in ECG by improving the detrending method that was originally proposed for eliminating slow non-stationary trends from heart rate variability (HRV). In our proposed method, a global trend is estimated in terms of baseline wandering by merging the local trend based on an ECG segment that represents a part of the ECG signal. The experimental results show that the improved detrending method can efficiently resolve baseline wandering without distorting any morphological characteristic embedded in the ECG signal in no time delay manner.