The invention discloses a method for eliminating thin layer tuning effect based on prestack gather. The method comprises the following steps: preprocessing data about prestack seismic gather of a target work area so as to correct an event and/or improve the signal-to-noise ratio of prestack seismic data; carrying out wavelet scale decomposition on the frequency spectrum of the preprocessed prestack seismic data so as to obtain a plurality of wavelets of different scales and analyzing and determining to-be-suppressed wavelet scales; reconstructing the wavelets of different scales by constructing scale bound terms so as to weaken information of the to-be-suppressed wavelet scales; and forming detuned seismic gather through of the reconstructed wavelets and the prestack seismic data. According to the invention, the targeted method for eliminating thin layer tuning effect based on wavelet scale decomposition is fundamentally brought forward by researching the generation mechanism of the tuning effect of prestack seismic gather; and the accuracy of earthquake prediction on a thin reservoir (or thin hydrocarbon reservoir) is improved by suppressing large-scale information of the seismic data to recover actual AVO characteristics caused by physical property parameters of the thin reservoir.