Research Article
LWTP: An Improved Automatic Image Annotation Method Based on Image Segmentation
@INPROCEEDINGS{10.1007/978-3-030-00916-8_6, author={Jianwei Niu and Shijie Li and Shasha Mo and Jun Ma}, title={LWTP: An Improved Automatic Image Annotation Method Based on Image Segmentation}, proceedings={Collaborative Computing: Networking, Applications and Worksharing. 13th International Conference, CollaborateCom 2017, Edinburgh, UK, December 11--13, 2017, Proceedings}, proceedings_a={COLLABORATECOM}, year={2018}, month={10}, keywords={Image annotation Tag propagation Image segmentation Local weights}, doi={10.1007/978-3-030-00916-8_6} }- Jianwei Niu
Shijie Li
Shasha Mo
Jun Ma
Year: 2018
LWTP: An Improved Automatic Image Annotation Method Based on Image Segmentation
COLLABORATECOM
Springer
DOI: 10.1007/978-3-030-00916-8_6
Abstract
Automatic image annotation is a technique that can be used to quickly generate tags for a massive dataset based on the content of the images. Nearest-neighbor-based methods such as TagProp are successful methods which have been used for image annotation. However, these methods focus more on weights based on the distances between the images and their neighbors, and ignore the weights of the different labels which can co-occur in the same image. In this paper, an improved method is proposed for automatic semantic annotation of images, which tags rare labels more effectively by processing the label matrix of the training set. In addition, image segmentation and data-driven methods are adopted to provide differential weights to the tags in one image, to improve the accuracy of the predicted tags. Experimental results show that the proposed method outperforms many classical baseline methods and can generate better annotation results than state-of-the-art nearest-neighbor based methods.
