A Trace transform based on subspace method for face recognition

A novel combination of Trace transform and subspace method is proposed for face recognition in this paper. This method is a general framework which can be used in most tasks of images classification. We extract a number of characteristic features from facial images through taking Trace transform over a set of angular directions. By using different Trace functionals we can get different features. The Trace transform features are projected into a lower dimensional facial subspace using linearity distinction analysis. And we use nearest neighbor rule and Euclidean distance to classify. The experimental results demonstrate that the proposed method has better average recognition rate than PCA, LDA and Gabor feathers, and has strong robustness to noise.