Landscape regionalization plays an important role in delineating the heterogeneous characteristics of territory, and provide the spatial fundamental data for natural resource planning and environmental protection activities. The integrating of the diversity indices (landscape metrics) is expressed the change of landscape structure by the richness and evenness of land-use objectives. In this study, a quantitative landscape regionalization framework is designed from 03 group (attribute factor, driving factor, and diversity factor) of basic landscape unit. By using k-means clustering, the study is classified into 06 sub-regions of 68 watersheds in the administration boundary of Van Chan district, Yen Bai province. With the comparison of region numbers in statistical and practical dimensions, the optimal results are edited and determined 15 sub-regions for uncertainty reduction of landscape regionalization.