The two-dimensional particle image velocimetry (PIV) data are inevitably contaminated by noise due to various imperfections in instrumentation or algorithm, based on which the well-established vortex identification methods often yield noise or incomplete vortex structure with a jagged boundary. To make up this deficiency, a novel method was proposed in this paper and the efficiency of the new method was demonstrated by its applications in extracting the twodimensional spanwise vortex structures from 2D PIV data in open-channel flows. The new method takes up a single vortex structure by combining model matching and vorticity filtering, and successfully locates the vortex core and draws a streamlined vortex boundary. The new method shows promise as being more effective than commonly used schemes in open-channel flow applications.