Object category pose estimation based on sparse representation and rank minimization

This paper presents a novel framework for object category pose estimation. The novelty of our approach consists in combining the low rank and sparse representation with the PCA-HOG feature[1] so as to estimate the object pose quickly and accurately. Moreover, the refinement mechanism formed by discriminant method is integrated to improve the performance of the classification between the opposite pose. We evaluate our approach for the class car on 3D categories dataset and the EPFL car dataset. Experimental results show that our method outperforms the state-of-the-art.