Seismic data with enhanced resolution allow interpreters to effectively delineate and interpret architectural components of stratigraphically thin geologic features. We used a recently developed time-frequency domain deconvolution method to spectrally balance nonstationary seismic data. The method was based on polynomial fitting of seismic wavelet magnitude spectra. The deconvolution increased the spectral bandwidth but did not amplify random noise. We compared our new spectral modeling algorithm with existing time-variant spectral-whitening and inverse -filtering algorithms using a 3D offshore survey acquired over Bohai Gulf, China. We mapped these improvements spatially using a suite of 3D volumetric coherence, energy, curvature, and frequency attributes. The resulting images displayed improved lateral resolution of channel edges and fault edges with few, if any artifacts associated with amplification of random noise.