General image classification based on sparse representation

Sparse representation based classification algorithm has been used to solve the problem of human face recognition. The image database is confined to human frontal faces with only illumination and slight expression changes. Cropping and normalization of the face need to be done in advance. In this paper, we apply the sparse representation based algorithm to the problem of general image classification, with a certain degree of intra-class variations and background clutter. Experiments have been done with the sparse representation based algorithm and SVM classifiers on 25 object categories selected from Caltech101 dataset. Experimental results show that without the time-consuming parameter optimization, the sparse representation based algorithm achieves comparable performance with SVM. We argue that the sparse representation based algorithm can also be applied to general image classification task when appropriate image feature is used.