Abstract: An Arctic ice layer reverberation suppression method combining seeded region growing and multidimensional property differences is proposed. Based on the analysis of the influence of the variations of the ice-layer reflection and scattering coefficients on the frequency-domain characteristics of reverberation signals, the reverberation suppression problem is transformed into the feature description and application of multidimensional property differences between it and the target echo signals. The method improves the seeded region growing method by employing a joint approximate entropy-Pearson correlation coefficient feature to describe the stability and similarity of one-dimensional instantaneous frequency sequences between reverberation and target echoes, thereby localizing echo positions. Combined with Rényi entropy to characterize the information distribution in the two-dimensional time-frequency plane, the region growing process is optimized to suppress adaptive reverberation in ice layers. The results of simulations and real Arctic data demonstrate that the method can effectively suppress reverberation and accurately extract target echoes in various ice layer environments, enhancing active sonar target detection in Arctic environments.