Images projecting on non-planar surfaces result in the geometric distortion which is highly undesired in the projection of digital content. Geometric correction techniques pre-warps the image to be projected such that the projected image appears geometrically correct for a given view on the non-planar surface. In this paper, we propose a piecewise planar geometric correction approach for a calibrated Projector Kinect System based on Kinect's point cloud data. A novel point cloud guided geometric correction (PGGC) technique is proposed which uses calibration parameters along with point cloud data to compute compensating images. The proposed PGGC method incorporates a novel sampling scheme based on the planar region extraction. Our sampling scheme selects a lesser number of samples in the regions corresponding to planar patches compared to regions corresponding to curved regions and edges. This results in the reduction in the number of sample points thereby reducing the computations. A novel brightness correction technique is applied to compensate for non-uniform brightness. Experimental results using different non-planar surfaces with different viewpoints show that the proposed approach achieves lesser distortion in comparison with state of the art techniques.
