Gabor face clustering using affinity propagation and structural similarity index
- PDF / 854,760 Bytes
- 9 Pages / 439.37 x 666.142 pts Page_size
- 37 Downloads / 224 Views
Gabor face clustering using affinity propagation and structural similarity index Issam Dagher 1
1
& Sandy Mikhael & Oubaida Al-Khalil
1
Received: 16 March 2020 / Revised: 31 July 2020 / Accepted: 2 September 2020 # Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract
Clustering is an important technique in data mining. It separates data points into different groups or clusters in such a way that objects in the same group are more similar to each other in some sense than with the objects in other groups. Gabor face clustering using affinity propagation and structural similarity index is composed of: A representation based on Gabor filters which has been shown to perform very well in face features, Affinity propagation clustering algorithm which is flexible, high speed, and does not require to specify the number of clusters, and structural similarity index which is a very powerful method for measuring the similarity between two images. Experimental results on two benchmark face datasets (LFW and IJB-B) show that our method outperforms well known clustering algorithms such as kmeans, spectral clustering and Agglomerative. Keywords Face clustering . Affinity propagation . Gabor . Structural similarity index
1 Introduction Face clustering problem has found many applications nowadays in many emerging areas like forensic and surveillance applications, smart phones, and social media. It can be described as follows: Given a set of data or faces, clustering techniques separate them into different group. Each group shares some common characteristics (faces of the same person). The greater the similarity within the group (same person) and the greater the difference between groups (different persons), the better the clustering is. This problem can be tackled by different approaches. Different face representations can be considered and different cluster algorithms can be applied. Good model should be used to represent a face and good cluster technique with powerful similarity measure should be be applied on this model in order to get the best possible clustering of a set of faces.
* Issam Dagher [email protected]
1
Computer Engineering Department, University of Balamand, Tripoli, Lebanon
Multimedia Tools and Applications
1.1 Related work In this paper, we have attempted to handle the face clustering problem by using a set of Gabor filters as a face representation method and by using an effective face clustering algorithm (Affinity propagation) which is capable of automatically grouping the faces. We replaced the popular Euclidian similarity measure by a more powerful one which is the Structure Similarity Measure. Different clustering techniques use different similarity measures. K-means [19] algorithm is the simplest clustering algorithm. It minimizes the within-cluster sum of squares (Euclidian similarity measure). The number of clusters should be specified. Hierarchical clustering algorithms fall into 2 categories: top-down or bottom-up. Bottom-up algorithms treat each data point as a
Data Loading...
