Incorporate Spatial Information into pLSA for Scene Classification

Abstract

pLSA has been successfully used in scene classification as an intermediate representation of images, but it didn't utilize the spatial information of an image which is important for scene classification tasks. To improve the accuracy of classification, we proposed a new method which incorporates spatial information coming from neighbor words and topics' position into pLSA. Finally, an image can be represented by the position distribution of each latent topic, and subsequently, we train a classifier on the topics' position distribution vector for each image. Besides, the traditional fold-in heuristic way of pLSA is not necessary and more sophisticated supervised pLSA can be adopted when our no-fold-in way is used, whichalso givesan accuracy improvement. © Springer-Verlag Berlin Heidelberg 2013.