Scene image classification using locality-constrained linear coding based on histogram intersection

Recently linear Spatial Pyramid Matching (SPM) method based on sparse coding has achieved great success in image classification. The raise of Locality-constrained Linear Coding (LLC) proves the importance of locality. In this paper, we propose an improved feature coding scheme called Locality-constrained Linear Coding Based on Histogram Intersection (HILLC). HILLC uses histogram intersection to describe the distance between feature vector and codebook. For each feature vector, search the KNN nearest neighbors to construct a local codebook. Compared with LLC, HILLC can obtain more robust codes. Experimental results demonstrate that our proposed method outperforms other related coding methods.