Object Detection in Images With Cluttered Background by Using Local Features and Their Configuration
In computer vision object detection is dealing with the problem to find certain objects in an image. Under controlled conditions the results of current systems are reliable. The absence of background clutter is such a condition. Problems arise if these conditions do not hold. In this paper we propose a method for overcoming the problem of background clutter by using a sensitive voting for objects and taking into account the position of local features. In an evaluation our proposed method clearly outperforms a naive object voting, by returning for 64% of the images the correct object compared to 4%.