Abstract: The analysis and interpretation of ground-penetrating radar (GPR) echo signals mostly rely on the experience of professionals and are thus highly subjective, affecting detection efficiency and accuracy. To address the challenge of poor imaging quality induced by the indistinct cavity signal features after rebar signal suppression, this study proposes a layered frequency-domain range migration imaging algorithm that combines deconvolution filtering and curvelet transform. Specifically, to compress the side lobes of wavelets and enhance the characteristics of weak reflected signals, deconvolution filtering is incorporated into data preprocessing, thus focusing the reflected energy along cavity boundaries. To retain the details of curve edges and improve imaging quality, the curvelet transform is employed. The proposed algorithm, which combines both, allows for the effective characterization of the geometric features of cavity targets while highlighting the dominant frequency components of signals, thereby enhancing imaging quality.