The reflection of thin layer may lead to frequency notch and the reciprocal of notch cycle is equal to the time thickness of thin layer. The frequency spectrum decomposition technology takes this as the rationale to predict the layer thickness. Starting from this rule, the high fidelity requirement of original seismic data and principles defining frequency range acquired by frequency spectrum decomposition was elaborated. In actual application, tuning cubes along interest horizons was used to predict layer thickness through calculating the first maximum frequency in frequency domain; the single frequency tuning cube was used to detect the thin beds; the changes of tuning energy was used to describe the spatial distribution characteristics of lateral disconnected geologic abnormal bodies. The application results show that the spectral decomposition technology has a greater ability of recognizing geologic bodies than the conventional method which is based on main frequency in time domain.