Spectral decomposition is a proven tool in seismic interpretation, aiding interpreters to highlight channels, map temporal bed thickness and other geological discontinuities. Once seismic data is spectrally decomposed, notch patterns in the amplitude spectra are indicative of the reservoir layers thickness and/or its interval velocity. Additional cepstral decomposition will allow direct extraction of bed time-thickness or arrival time under particular reflectivity series setup. We build on these observations to establish a more generalized workflow for reflectivity retrieval method without the need to understand the details of the wavelet, provided the starting seismic is stably phased via phase correction during processing. We demonstrate reflector time and its apparent strength can be identified in a transformed seismic resonance domain resulted from a modified cepstrum analysis. In this domain, each reflector can be characterized from obvious linear hot spots. The timing and strength of those linear hot spots will reveal reflector times and scaled reflectivity coefficients. This new method is subsequently applied for thickness prediction of a target reservoir in a complex geological setting, with large thickness variations and weak impedance contrast with underlying lithology previously complicating identification of base-reservoir. In a deep-water field blind test, the sand thicknesses evaluated from this method are found to be close to true vertical thickness found in wells.