Text line segmentation in Chinese handwritten text images
Chengdong Zhang, Guee-Sang Lee · 2011
In this paper, text line segmentation based on 2-D Tensor Voting is proposed, 2-D tensor voting is used to remove outliers and located center points of connected components. The saliency values and direction of normal vectors are represented information of tensors for segmentation. The text images obtained from HIT-MW dataset.