As an important part of evaluating the performance of glass-ceramic materials, bubble defect detection can guide the cutting and reprocessing of the original glass-ceramics to give full play to its stable optical characteristics. Due to the problem that the background brightness of the field of view is uneven and the outer periphery of the glass-ceramic is brightened when the circular light source illuminates the cylindrical glass-ceramic, a glass-ceramic bubble detection algorithm based on grayscale and geometric characteristics is introduced. Possible bubble outlines are obtained by Canny edge detection according to the local image gray feature, and then the false outlines are removed according to bubble geometric characteristics to obtain the real bubble defect information. Experimental results show that the algorithm can effectively extract bubble defects and improve the accuracy of bubble defect detection.