Real-Time Detection of Landscape Scenes

In this paper we study different approaches that can be used in recognizing landscape scenes. The primary goal has been to find an accurate but still computationally light solution capable of real-time operation. Recognizing landscape images can be thought of a special case of scene classification. Even though there exist a number of different approaches concerning scene classification, there are no other previous works that try to classify images into such high level categories as landscape and non-landscape. This study shows that a global texture-based approach outperforms other more complex methods in the landscape image recognition problem. Furthermore, the results obtained indicate that the computational cost of the method relying on Local Binary Pattern representation is low enough for real-time systems.