Glacial lake is one of the most important sensitive factors that reflect the global climate change. The accurate extraction of glacial lake information can be used to analyze the status of glacial lake in different period,and to provide the possibility for the analysis of climate change.Focus on the complex spectra characteristics of glacial lake,and considering that the C-V model has a complete theoretical basis,but the huge amount of calculation,so this paper combines the C-V model with the symbolic pressure function and the "global to local Iterative water extraction algorithm to study the a new method of glacial lake extraction in High Asia. This paper chooses the three typical development region of glacial lake, Himalaya mountains,Southeastern Tibet,Altai mountains.Glacial lake were extracted by using Landsat 8 OLI image, and then this paper do the research of the accuracy assessment of the glacial lake extraction results. The C-V model has better anti noise performance and weak edge segmentation effect,so the experiment obtained good results. The result shows that overall accuracy of glacial lake extraction is 89.86 %, and the method can extract the glacial lake's information rapidly and accurately.