<p>Mesoscale eddies (MEs) are significantly important time-dependent swirling mechanism of seawater movement, influencing material transport and energy exchange at the global scale. Automatic identification of MEs is a key step for tracking their evolution and exploring their interactions with other ocean phenomena. Currently, satellite altimeter datasets are widely used in automatic identification methods. Existing approaches are categorized as parameter-intensive or low-parameter (parameter-efficient) methods, which suffer from problems such as high sensitivity, parameter tuning, and limited applicability to multi-core eddies. To address these issues, in this study, a progressive shrinking strategy involving minimal parameter requirements was proposed for identifying MEs. Specifically, MEs were treated as connected pixel regions, and Sea Level Anomaly (SLA) threshold-data were developed to generate binary images. The threshold was initially set to zero and was automatically updated throughout the process. Connected regions in the binary images were then labeled using a connected-component algorithm and evaluated according to eddy shape and size constraints. The proposed method was validated across four specific ocean regions: the Northwest Pacific, Southeast Pacific, South Atlantic Ocean, and South China Sea. Results show that, compared with the traditional SLA-based method, the proposed approach more effectively captures multi-core structures with diverse characteristics while maintaining low parameter dependence.</p>