Scene Classification in Images B V V Sri

In this paper, following the works of Oliva and Torralba, we present a procedure to classify real world scenes in eight semantic groups of coast, forest, mountain, open country, street, tall building, highway and inside city, without going through the stages of segmentation and processing of individual objects or regions. The approach essentially takes into account the diagonistic information stored in the power spectrum of each category of images and through supervised learning, separates characteristic feature vectors of each class in separate groups that helps assign a new test image to its respective group. We follow a sequential hierarchy in which images are first sorted according to their naturalness and through traversing a tree, we ultimately reach the desired node that represents the class of the image. The significance of results obtained and the ripple effect of errors have also been reflected upon.