Robust indoor/outdoor scene classification

With the exponential growth of storage of digital images, retrieval has become an impending issue. Such large collection of data takes a considerable amount of time in retrieving images apart from picking relevant images with respect to the query. Despite advancements in introducing effective features, the search time still remains larger. In such scenario the search time could be minimized by categorizing the database scenes into indoor or outdoor. The objective of the paper is to categorize an image into indoor or outdoor scene. To support automatic scene classification at the concept level, an efficient illumination and rotation invariant low level features such as color from HSV color model and texture (GBWHGOPCA) features have been used in conjunction with Sparse Representative Classifier (SRC). Since these image features exhibit a distinctive disparity between images containing indoor or outdoor scenes, the proposed method achieves better performance in terms of classification accuracy. This work is evaluated on IITM-SCID2 (scene classification image database) and 15 scene category dataset and dataset of 3442 images collected from the web by authors.