Seismic characterisation of carbonate reservoirs is a challenging task because of special geological and geometrical properties of the reservoirs. This challenge is also due to the low signal-to-noise ratio and low resolution of seismic data. We improved the data quality by applying an inverse-Q filter to 3D seismic data, and in turn increased the accuracy of reservoir characterisation. We implemented the inverse-Q filter in a stabilised fashion so that we can compensate seismic amplitudes and correct for waveform phases simultaneously. We conducted the case study on 3D seismic data from the Ordovician carbonate reservoirs in the Tarim Basin, China, where the Ordovician sediment is buried very deep, with a depth larger than 5000 m, and has extremely strong heterogeneity either in the vertical or the horizontal direction. The stabilised inverse-Q filter improved the lateral continuity of seismic reflections, enhanced the spatial and vertical resolution, and strengthened weak signals in 3D seismic data. Therefore, we were able to characterise the interior details of the unconformity of the Ordovician sediment and to illustrate vug caves and fault fractures within the target carbonate reservoirs.