Fast Semi-supervised Segmentation of in Situ Tree Color Images

摘要: In this paper we present an original semi-supervised method for the segmentation of in situ tree color images which combines quantization, adaptive fragmentation learning areas defined by human operator and labeling propagation. A mathematical morphology post-processing is introduced to emphasize narrow thin structures characterize branches. Applied L*a*b* system, well adapted easily adjust set so that resultant corresponds accuracy achieved operator. The has been embarked evaluated on a tablet help professionals their expertise or diagnosis. images, acquired processed with mobile device, more less complex background both terms content lightness, dense foliage thick Results are good soft lightness without direct sunlight.