Research on Image Segmentation Method of Ancient Text Rubbing Drilling Traces based on U-net Network
Year: 2024
Type: conference-paper
Abstract: Aiming at the problems of high noise and severe fracture damage in ancient text rubbings, a method of ancient text rubbings drilling trace segmentation based on u-net network is proposed. Firstly, the feature maps of various scales in the encoder are fused with the Mutle fuse block, which enables cross layer transmission of information. In the decoder section, the original size of the feature map is gradually restored through up sampling and deconvolution operations. Then, Afterwards, each level is fused with the corresponding feature map in the encoder through skip connections using the Mutle fuse block. This enables the U-Net network to achieve feature information fusion at multiple scales, improving its performance in drilling trace segmentation. To verify the effectiveness of the method, it was validated on a dataset of ancient text rubbings. The experimental results showed that the algorithm proposed in this paper achieved good results in the segmentation of ancient text rubbings images.
Authors: Xiaosong Shi, Yawen Huang, Yongge Liu
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: gold
OpenAlex ID: W4413080090