Abstract:

To overcome the deficiencies of low-rank embedding in data reconstruction and noise suppression, and improve the accuracy of its feature recognition,an image feature extraction algorithm is proposed based on orthogonal projection learning. The half-quadratic alternating direction method of multipliers algorithm is designed to solve the orthogonal projection learning model. The model retains the main features of the samples by introducing an orthogonal matrix,the norm constraints makes the extracted features more prominent, and the weighted Schatten p-Norm is used to approximate the optimal solution of the rank. To improve the robustness of the model and make it suitable for supervised scenarios, generalized correntropy is used for data item modeling and classification loss function construction. Experimental results on different scale datasets show that the proposed model has better feature extraction performance than other existing models.

Key words: image feature extraction, weighted Schatten p-norm, lowrank representation, projection learning