The three-dimensional high efficiency video coding (3D-HEVC) is an emerging standard to compress multi-view video plus depth (MVD) data. In 3D-HEVC, many new approaches are developed for depth intra coding. In addition to the 35 intra prediction modes inherited from HEVC, depth modeling modes (DMMs) are also employed to better preserve the object edges, which improve the coding efficiency significantly, but at the same time put heavy computational burden on the encoder. To reduce the complexity, we propose a fast depth intra coding scheme using gray-level co-occurrence matrix (GLCM) to classify the coding units (CUs) into different types, and avoid checking some unnecessary modes for certain CUs. Experimental results show that the proposed algorithm can achieve on average 19.16% time reduction, with a negligible Bjontegaard Delta bitrate (BD-rate) increase of 0.12% on synthesized views, compared with the original 3D-HEVC test model HTM-15.0.