Auto Center Find Density Peaks Algorithm and its Application in Face Image Clustering

Face image clustering is one of the most important applications in digital signal process. Here, Find Density Peaks clustering algorithm with auto choosing centers is proposed and applied into the face image clustering problem. In the clustering process, the Complex Wavelet structural Similarity Index is introduced to quantize the difference between two face images. The deciding mechanism of two key parameters are modified. Avoiding the false centers, the true centers of each cluster are auto defined. And the face images dataset are properly clustered. The face images are dimensionality reduced and visualized via T-SNE as scatter points in 2D space. The results show a good partition of face images dataset. And the proposed algorithm shows practical values in most applications required with clustering.