Ebullition is the primary pathway for methane (CH₄) emissions from freshwater ecosystems such as lakes and reservoirs, yet it always exhibits significant spatiotemporal variability. In shallow zones near reservoir inlets, CH₄ is often generated in the sediments and released as bubbles. However, capturing the ebullition hotspots in open water bodies remains a challenge. During the ice-covered season, bubbles rich in CH₄ are trapped and preserved as “ice-encased bubbles” within the growing ice layer. This natural phenomenon provides an opportunity to identify CH₄ ebullition hotspots based on the characteristics and spatial patterns of ice bubbles. This study focuses on the Dongfeng Reservoir, a large reservoir in northeastern China. Using high-resolution unmanned aerial vehicle (UAV) imagery of the winter ice surface and in-situ measurements of ice bubbles, we developed an object-based image analysis method for bubble segmentation and extraction. Spatial autocorrelation analysis was employed to investigate the distribution patterns of ice bubbles and to identify CH₄ ebullition hotspot areas. The important findings include: (1) The diameters of ice bubbles in Dongfeng Reservoir ranged from 1 cm to 10 cm. UAV imagery captured at 15 m altitude achieved a spatial resolution of 0.4 cm, sufficient for high-quality bubble identification. A morphological top-hat transformation effectively corrected uneven illumination and enhanced the quality of digital orthophoto mosaics. (2) Differences in shape attributes, brightness, and density allowed for effective classification of bubbles and ice cracks, with the overall classification accuracy exceeding 0.8 across three investigated regions, demonstrating robust extraction performance. (3) The total bubble area in the three dense bubble regions accounted for 0.24% of the reservoir inlet area. Within these regions, the bubble area ratio along ten surveyed transects ranged from 2.6% to 7.8%. Spatial autocorrelation analysis revealed significant clustering of ice-encased bubbles, with CH₄ ebullition hotspots occupying only 0.9% of the reservoir inlet area. The results indicate that CH₄ ebullition may be concentrated in the limited regions across the reservoir. These findings provide important insights into the spatial distribution characteristics of methane emission hotspots in lakes and reservoirs.