Supervoxel Approach in an Efficient Multiscale Object-Based Framework

• Lucca S. P. Lacerda PUC Minas
• Felipe D. Cunha PUC Minas
• Silvio Jamil F. Guimarães PUC Minas

DOI: https://doi.org/10.5753/sibgrapi.est.2023.27469

Resumo

The use of supervoxel segmentation has shown substantial improvement in video analysis because it can improve object delineation and reduce computer workload. In this work, we have used SICLE (Superpixel Through Iterative Clearcutting), which is an innovative graph-based superpixel framework that makes multi-scale segmentation by exploiting object information. For segmenting videos, we changed the graph creation step. The framework has exceeded state-of-the-art approaches, and its results precisely delineate the object of interest.

Referências

B. Wang, Y. Chen, W. Liu, J. Qin, Y. Du, G. Han, and S. He, “Real-time hierarchical supervoxel segmentation via a minimum spanning tree,” IEEE Trans. Image Process., vol. 29, pp. 9665–9677, 2020. [Online]. Available: https://doi.org/10.1109/TIP.2020.3030502

F. Belém, B. Perret, J. Cousty, S. J. F. Guimarães, and A. X. Falcão, “Efficient multiscale object-based superpixel framework,” CoRR, vol. abs/2204.03533, 2022. [Online]. Available: https://doi.org/10.48550/arXiv.2204.03533

F. de Castro Belém, B. Perret, J. Cousty, S. J. F. Guimarães, and A. X. Falcão, “Towards a simple and efficient object-based superpixel delineation framework,” in 34th SIBGRAPI Conference on Graphics, Patterns and Images, SIBGRAPI 2021, Gramado, Rio Grande do Sul, Brazil, October 18-22, 2021. IEEE, 2021, pp. 346–353. [Online]. Available: https://doi.org/10.1109/SIBGRAPI54419.2021.00054

C. S. J. de Almeida, F. Belém, S. A. Carneiro, Z. K. G. do Patrocínio, L. Najman, A. X. Falcão, and S. J. F. Guimarães, “Graph-based supervoxel computation from iterative spanning forest,” in Discrete Geometry and Mathematical Morphology - First International Joint Conference, DGMM 2021, Uppsala, Sweden, May 24-27, 2021, Proceedings, ser. Lecture Notes in Computer Science, J. Lindblad, F. Malmberg, and N. Sladoje, Eds., vol. 12708. Springer, 2021, pp. 404–415. [Online]. Available: https://doi.org/10.1007/978-3-030-76657-3_29