Image clustering using local discriminant model and two-dimensional PCA features

摘要

Recently, local learning based image clustering model was proposed that utilized discriminant analysis. In local discriminant model and global integration (LDMGI) model, local discriminant model was developed to evaluate image clustering at local level, and the optimal image features were obtained using image interpolation approach. We performed further image feature reduction through two-dimensional PCA (2DPCA) by extracting significant eigenvectors of the image dataset. Because, by projecting image features in principal component analysis (PCA) space, we can remove principal components of scatter matrices with small eigenvalues. Due to which, LDMGI model is more effective and efficient. We evaluated the performance of proposed 2DPCA-LDMGI image clustering model using 10 benchmark image datasets, and report significant overall performance improvement over previous LDMGI model. Further, 2DPCA-LDMGI is computationally efficient on all image datasets and overall computational cost is reduced to more than half as compared with LDMGI model.