An Efficient Parallel Approach for 3D Point Cloud Image Segmentation Using OpenMP

The 3D Point cloud segmentation process is used in Image processing to detect the edges and the 3D structure and widely used in the computer vision and robotics. The 3D point cloud segmentation consists of three major processes: Point cloud data fitting on the 3D Array (Grid), Gradient calculation and thinning process. This process is computationally expensive due to many multiplication and addition which are required to calculate the gradient for the identification of edges, to find the location where a point to be located and stored and thinning process. This is why the existing algorithms are very slow to run on a single processor sequential programming. Therefore it is necessary to make it parallel for the high performance and to speed-up the computation. In this study, a parallel processing approach was used for the segmentation of 3D Point cloud image. The proposed method uses a 3D Vector structure grid concept for the creation of Virtual Grid to store the huge number of unordered points for the fast processing in order to find the neighborhood points. Our algorithm focuses on the fast extraction of edges by segmentation using gradient 3D sobel operator in parallel approach. The results show that the parallel approach will be efficient and provides a better performance in finding the edges. We evaluated this approach with 3 artificially generated data sets in two implementations: one in sequential and the other one in parallel. This new approach produces improved speedup in the edge detections process.