The time-lapse seismic survey for sequestration study at Cranfield can be problematic because of misalignments between time-lapse data sets. Such misalignments can be caused by the seismic data processing workflow and may result in the wrong interpretation of time-lapse seismic amplitude differences. We propose an efficient local-correlation-based warping method of registering the time-lapse poststack data sets, which can align these data sets without changing original amplitudes. Application of our registration method to Cranfield time-lapse data demonstrates its effectiveness in separating time-shift character from seismic amplitude signature. After registration, time-lapse differences show an improved consistency in vertical cross sections and a more localized distribution of difference amplitudes along the horizon, allowing us to apply a high-resolution basis pursuit inversion (BPI) for acoustic impedances. Inversion results show that decreases in acoustic impedances occur mostly at the top of the injection interval, which can be used as an indicator of rock properties to detect a subsurface plume.