Abstract: To address the low accuracy of reservoir characterization in XiHu Sag in a coal-rich environment, this study developed a matching pursuit technology based on AVO information constraints combined with the AVO intercept and gradient characteristics of coal. It can suppress the strong reflectance lithologic artifacts caused by coal and highlight actual and effective reservoir signals. Based on the negative intercept <i<P</i< and positive gradient <i<G</i< of the AVO of coal, the seismic–sensitive factor <i<P</i<–<i<G</i< of coal identification was developed to amplify the seismic response of coal and suppress the high-amplitude response of non-coal. Then, to accurately identify the location of coal, the seismic information of coal was used as the original signal that needs to be decomposed and reconstructed by matching pursuit. Additionally, the efficiency of signal-matching decomposition was improved using the technology of complex seismic track analysis. Finally, the strong reflection elimination of coal was completed. Model trials and practical applications indicate that this method could accurately identify the seismic response location of coal and improve the efficiency of the matching pursuit algorithm. Moreover, the coal-eliminated seismic data can better highlight the lateral distribution changes of the reservoir and improve the vertical characterization accuracy of the main gas layer.