Image segmentation based on transmission characteristics of flue-cured tobacco leaves

Reflection and transmission images of flue-cured tobacco leaf were both taken in specially-designed lamp box, and then imported into computer for digital analysis. The comparison between the tobacco leaf area and background area for RGB attributes in these two types of images were conducted by sampling with a moving 3×3 matrix of pixels. It was found that most of the blue vector value of pixels in transmission image of tobacco leaf were equal to zero and few of them with greater than zero were distributed discontinuously within the matrix. It was also found that the relative standard deviation (RSD) of blue vector value in a matrix with nine pixels located on the tobacco leaf area in the transmission image was either 0 or more than 100%. Based on the findings, an approach to image segmentation with area-oriented statistics was thus proposed in this paper. The results show that the segmentation accuracy is higher than that by the method of point-oriented statistics, and the average pixels of all segmented image in this study were 0.65% more than those by point statistics per tobacco leaf. The characteristics of the original object image were well kept after segmentation. The presented method may be helpful to increase segmentation accuracy and lay foundations for further studies by using transmission characteristics of tobacco leaf.