Concurrent Image Classification and Annotation Using Efficient Multi-layer Group Sparse Coding

The multi-layer group sparse coding framework is presented for the purpose of image classification and annotation. This paper introduces the multi-layer group sparse structure of the image reconstruction coefficients to leverage the needs between the class label and tags. The sparse structure translates the mutual dependency among the class label that defines the whole image content. The tags define the components of the image content. Hence, the multi-layer tag propagation method involving the combination of the class label and subgroups with the similar tag distribution is utilized in the annotation process of the test images. The multi-layer Group sparse coding is extended in the Reproducing Kernel Hilbert Space (RKHS), so as to develop the proposed methodology to become more appropriate for nonlinear separable features and enhance the overall efficiency of the image classification and annotation. The multi-layer Group sparse coding is integrated with the strategy that significantly improves the computational efficiency. From the experimental results, the proposed technique outperforms the standard techniques and achieves outstanding performance for the image classification and annotation