The ice formation generally traps air bubbles, with varying sizes and distributions within the ice matrix, which significantly affect its optical clarity, heat conduction, and physical properties in both natural and engineered systems. In this study, the features of the trapped bubbles are recorded by a camera in a horizontal Hele-Shaw cell at various freezing temperatures. A robust image processing framework is proposed for automated detection and quantitative analysis of entrapped air bubbles in ice samples formed over a regulated temperature range (-15°C to -35°C). The algorithm achieved a mean accuracy of 94.6% in bubble detection, consistently performing across a broad range of bubble sizes and challenging imaging conditions. In addition, compared to the deep learning approach, which struggles with small or overlapping bubbles, the proposed method achieves more reliable and precise detection. Temporal and spatial analyses revealed that lower freezing temperatures greatly enhanced bubble nucleation rates, along with morphological parameters like length, width, and aspect ratio, consistent with the conventional manual technique. These findings offer new insights into ice microstructure formation and highlight their importance for industrial freezing processes, where controlled air entrapment influences key product quality features.
