Synchronous Segmentation of Reflection and Transmission Images of Flue-Cured Tobacco Leaves

摘要

In this study, the reflection and transmission images of a whole leaf were acquired separately without shifting its position using a lamp box device. Based on the intrinsic transmission characteristics of flue-cured tobacco leaves, an image segmentation method was proposed to achieve the synchronous background correction of the reflection and transmission images of one same flue-cured tobacco leaf. By combining the point-oriented statistical segmentation method and the area-oriented statistical segmentation method, this method achieved the goal of acquiring the pure images of paired leaves with identical effective pixels in contours. It is proved that this combined method achieves a higher segmentation precision than the simple adoption of the point-oriented statistical segmentation or the area-oriented statistical segmentation method, and the boundaries are smooth and continuous. All these have laid a solid foundation for the subsequent digital tobacco grading based on both reflection and transmission images.