Poverty is a challenge worldwide. Policy and regulations guiding anti-poverty measures for governments, NGOs, and multilateral institutions have not considered the spatial scale effect of regional poverty, resulting in low-efficiency poverty alleviation actions. This study addressed research gaps by analyzing the multi-scale (county, township, and village) features of regional poverty in Yanbian Korean Autonomous Prefecture in Jilin province, China. It examined the impact of geographic capital and associated spatial heterogeneity from four dimensions: natural environment, transport location, facilities accessibility, and socioeconomic development. The results identified that regional poverty varied at different scales: lower-scale poverty had higher levels of spatial differences, agglomeration, and spatial autocorrelation than higher-scale poverty, and the “island effect” was prominent. The factors potentially impacting regional poverty varied at different scales for geographical capital. At the township scale, only transport location and socioeconomic development dimensions could make significant differences. Factors in all four dimensions could affect village-scale poverty significantly, and the natural environment dimension was more effective than the other three dimensions. The impact of geographic capital and its spatial heterogeneity at the village scale varied, implying that local and diverse anti-poverty measures should increase. This study improves understanding of the multi-scale features of regional poverty and supports the formulation of effective anti-poverty measures.