Image low-level semantic feature extraction based on rough set
Shaoshuai Lei, Yun Gu, Cao Changqing, Xie Gang · 2012
Rough set theory can link classification and knowledge together. Therefore, rough set theory is applied to the image low-level semantic feature extraction in this paper. First, the decision table of low-level features is constructed, and then knowledge reduction of rough set is applied to reduce the decision table, which removes redundant samples and redundant attributes, and to identify effective semantic low-level features. Knowledge reduction can only deal with discrete data, therefore knowledge K-means clustering is used to normalize attribute decision table before knowledge reduction. Finally, we use support vector machine(SVM) to verify the validity of the extracted features. The experimental results show that the proposed method not only can guarantee the premise of image semantic recognition, but also greatly reduce the amount of computation.