Sparse code LBP and SIFT features together for scene categorization

Abstract

Local descriptors, Local Binary Pattern (LBP) and Scale Invariant Feature Transform (SIFT), are widely used in various computer applications. They emphasize different aspects of image contents. In this paper, we propose to sparse code them together for categorizing scene images. We first regularly extract LBP and SIFT features from the images. Then, corresponding to each feature, a visual word codebook is constructed by using training images. For creating a representation for an image, the LBP and SIFT features extracted from the same position of the image are encoded together based on sparse coding. Specifically, we combine the obtained LBP codebook and SIFT codebook as a two dimensional table. In this table, each entry corresponds to a LBP visual word and a SIFT visual word. Therefore, the encoding values of the entries depend on both features. After all features of the input image are processed, the spatial max pooling is adopted to determine the image representation. Obtained image representations are classified by utilizing SVM classifiers. Finally, we conduct extensive experiments on datasets scene categories 8 and MIT 67 indoor scene to evaluate the proposed method. Obtained results demonstrate that combining features in the proposed manner has improved the scene categorization performance significantly. In addition, the results of the proposed method are comparable to other state of the art works.