Multiple quadrilateral detection for projector-camera system applications

In this paper, a novel method to detect multiple quadrangles in images for the applications in projected reality, such as a smart office environment, is proposed. The proposed approach is able to detect rectangular surfaces with perspective distortion, which appear as quadrangles in 2D images, in a cluttered background accurately. It can deal with partial occlusion. A new technique called the Q-corner is devised based on randomized Hough transform (RHT). This makes the algorithm become effective and robust in finding multiple rectangular objects like white card-boards in the scene. We present the experiment results with synthetic and real data, showing that our method is more robust to noise than a standard contour-based algorithm and an existing line-based method. The promise of the method for potential applications has been demonstrated in a real scenario.