Abstract: In order to accurately inspect the cigarette capsules with bubble defects in release inspection and discriminate bubble defect from stains, white spots, and other minor surface defects, a high precision inspection method based on computer vision was proposed. Firstly, the capsule area was precisely extracted via the image segmentation in an HSV (Hue, Saturation, Value) color space; secondly, the LoG edge detection operator was improved to enhance bubble contour and suppress the contours of the other defects; finally, the extracted contours were coded according to the "double liquid ring" characteristics of the bubbles to accurately recognize the bubbles. The proposed method was tested with capsule samples of three different specifications and of different colors. The results showed that the average recognition rate of the proposed method for bubble defects was 98.9%, the average false positive rate was 0.21‰, and the deviation was ±0.03% comparing with manual inspection. The proposed method features a higher recognition rate for bubble defects, and provides a support for promoting the appearance quality inspection of cigarette capsules.