A Comparative Study of Two State-of-the-Art Feature Selection Algorithms for Texture-Based Pixel-Labeling Task of Ancient Documents.
Maroua MehriRamzi ChaiebKarim KaltiPierre HérouxRémy MullotNajoua Essoukri Ben AmaraPublished in: J. Imaging (2018)
Keyphrases
- feature selection algorithms
- pixel labeling
- feature selection
- probabilistic relaxation
- energy minimization
- information retrieval
- spatially variant
- data sets
- multi class
- feature set
- selection algorithm
- graph cuts
- semantic segmentation
- feature subset
- image processing
- object recognition
- classification accuracy
- learning models
- multiscale
- stability of feature selection