The characteristics of topographic relief variation are diverse from one landform type to another in the Loess Plateau. Great differences of topographic relief variation exist in different directions, which is known as topographic anisotropy. As fractal theory and geomorphology research become increasingly cooperative, Multifractal Detrended Fluctuation Analysis (MF-DFA) method, an effective means to quantitatively describe the nonlinear evolution process of complex sequences and the self-similar structural features formed by nonlinear evolution processes, manifests great potential in the study of topographic anisotropy of different landforms. In this paper, multifractal spectrum parameters were introduced to describe the variation characteristics of topographic anisotropy, in order to illustrate the local and global variation pattern of topographic anisotropy of different loess landforms. Based on 5 m-resolution Digital Elevation Model (DEM) data, three typical areas, Chunhua County of loess tableland, Ganquan County of loess ridge and Suide County of loess hill, were selected as the study areas. After that, MF-DFA of the variation characteristics of topographic anisotropy were conducted on these three areas. The results show that the topographic anisotropy of the three loess landform types has significant multifractal characteristics. The singularity strength of the topographic anisotropy from strong to weak turns out to be in the order of Chunhua, Suide and Ganquan counties. The probability of the singularity strength of topographic anisotropy in Chunhua County reaches the maximum at singular strength of 1.4, and there is a small peak around the singularity strength of 2.2. The probability of the singularity strength of topographic anisotropy in Ganquan County reaches the maximum at singularity strength of 0.7, with a second peak around the singularity strength of 1.8. Differently, the singularity strength of topographic anisotropy in Suide County is evenly distributed around 0.8, reaching the maximum at singularity strength of 0.8. The overall singularity strength of Chunhua and Ganquan counties are 1.4 and 0.7 respectively, while the probability distribution of topographic anisotropy in Suide County appears symmetrically. It can be concluded that the singularity strength of the topographic anisotropy is the highest in loess tableland, followed by that in loess hill. The singularity strength of the topographic anisotropy is the smallest in loess ridge. The results confirm that multifractal singularity strength is an effective method for the description and analysis of the topographic anisotropy of loess landforms, e.g., tableland, ridge and hill. This study can provide quantitative information support for the accurate classification of loess landforms as well as topographic indicators for regional water and soil erosion assessment.