A Synchronous Background Segmentation Method for theTransmission and Reflection Images of Tobacco Leaves

Objective This paper describes the design of a lamp box device that captures,from each flue-cured to-bacco leaf,transmission and reflection images in the same image position.According to the transmis-sion characteristics of tobacco leaves namely that the blue vector value in the leaf part of its RGBtransmission image is dramatically absorbed.Thus,this study proposes a novel segmentation methodto synchronously treat the transmission and reflection images of the same flue-cured tobacco leaf,combining it with a point-oriented and area-oriented statistical segmenting method to successfully ob-tain leaf parts in pairs which have the same effective pixel quantities and the same contours.The re-sults show that segmentation accuracy is higher than when only using point-oriented or area-orientedstatistical segmenting methods.Moreover,the boundary is smooth and continuous.These findingsare very helpful for digitally rating tobacco quality based on the combination of transmission and re-flection images.