Face Recognition Algorithm using LBPWT features
Amrita Biswas, Manojit Ghose · International journal of imaging and robotics · 2015
In this paper we proposea novel face recognition algorithm using the Local Binary Pattern Histogram Wavelet Features (LBPWT) a new image descriptor computed from discrete wavelet transforms of Local Binary Pattern histograms. The proposed features are computed over partitioned non overlapping blocks of the image. Besides being rotation invariant these features retain the discriminative nature of LBP histograms. For face recognition we fuse the proposed LBPWT features of the partitioned image blocks with the LBPWT features of the canny edge detected image. Experiments have shown that performance of the proposed algorithm based on LBPWT is much better than earlier versions of non invariant LBP, rotation invariant LBP and LBPHF features used in face recognition. For all the experiments we have considered only a single image for training since in most practical scenarios limited images are available for the training database. We have shown that even with a single training image the proposed algorithm provides encouraging results.