Two-dimensional wave spectra describe the distribution of wave energy in terms of frequency and direction. For unimodal seas (i.e., single peaked), wave spectra can describe the wave component accurately. However, they fail to describe wave components properly in more complex situations. The simultaneous existence of wind sea and one or more swell systems is a situation that occurs very often. Therefore, for convenient application, these wave components should be distinguished by partitioning, which often relies on the watershed algorithm. However, the watershed algorithm has some problems, such as oversegmentation, as its first step takes many small partitions and the elements of these small partitions cannot be separated in the subsequent steps. A new method called merging and expanding (ME) method that involves two main steps (i.e., element merge and partition expansion) based on the 2-D slope spectrum product of surface waves’ investigation and monitoring (SWIM) is presented in this article. The step called element merge and the principle of regional consistency (PORC) are adopted to avoid oversegmentation and to make partitions possess physical meanings while the step called partition expansion is implemented to decrease filter dependence. The result indicates that the ME method can suppress excessive segmentation and the partitioned wave parameters of the ME method are closer to that of WAVEWATCH III (WW3) data than SWIM products.