Data Dimension Reduction Based on Concept Lattices in Image Mining
Wang Li, Luo Wei · 2009
High-dimensional image feature data is an obstacle for image mining. In order to reduce the image feature data dimension, this paper introduces a method based on concept lattices. After introducing the basic concepts of formal context theory and the attribute reduction of concept lattices, the feature attribute set produce algorithm and the dimension reduction algorithm are put forward. In the algorithm, the image formal context, which is transformed from the HSV color feature, and then the attributes of the concept lattices are reduced. The experiment results prove that this method is efficient, and it outperforms the principal component analysis (PCA).