Landscape patterns have typically spatial heterogeneity and scale dependence; thus it needs to span multiple scales in landscape pattern analysis. Therefore, the spatial granularity is crucial in scale aggregation. In the hilly area of the Three Gorges Reservoir area, patches are fragmented, landscape complexity is significant, and landscape pattern characteristics and scale variation are still to be clarified. In this study, we focused on Zigui County in the Three Gorges Reservoir, set 23 grain size gradient levels within 1-400 m, quantitatively evaluated the landscape pattern index and fitting function of the county, towns, and small basins at different amplitudes to explore more margin space landscape pattern index characteristics associated with the change in particle size. Based on the grain size effect characteristics of the landscape index and the curve characteristics (maximum curvature point and extreme value point) of the fitting function, a suitable grain size threshold of the mountain landscape pattern index with different amplitudes was determined, so as to reveal the complexity and variability of mountain landscape structure in the hilly region of the Three Gorges Reservoir. The responses of different landscape indices to the changes in spatial granularity and spatial amplitude differed. The responses of different landscape indices to the changes in spatial granularity can be summarized as four trends:increase, decrease, fluctuation, and no obvious regular change. Patch Density and the Largest Shape Index are sensitive to changes in patch shape and size, whereas the sensitivity of Diversity Index to the change in grain size is inapparent. Patch Density, Edge Density, and Perimeter-Area Fractal Dimension are not sensitive to the change in spatial amplitude, whereas the Largest Patch Index, Landscape Shape Index, Interspersion and Juxtaposition Index, and Splitting Index are sensitive to the change in spatial amplitude and are suitable for prediction of the threshold of appropriate grain size for different amplitudes. Landscape indices such as Edge Density, Mean Patch Size, Landscape Shape Index, Contagion Index, and Percentage of Like Adjacency can be highly fitted to the curve function, and the curve characteristics of the fitting function can be used to identify the turning point of the change in the landscape index and reduce the difference in artificial judgment. Based on the changing trends and turning points of the landscape indices of 6072 raster data of 251 small basins, 12 townships, and Zigui County, the first-scale granularity threshold for landscape pattern analysis of small basins, towns, and Zigui County is 3 m, 4 m, and 7 m, while the second-scale particle size threshold is 50 m, 100 m, and 100 m, respectively, integrating the turning points of each landscape index. Thus, the smaller the spatial amplitude, the smaller the suitable grain size for landscape pattern analysis.