A Retinex theory based point sampling method

To accelerate the processing for integration, registration, representation and recognition of point clouds, it is of growing necessity to simplify the surface of 3-D models. This paper proposes a Retinex theory based points sampling method and the effectiveness of the sampling results are demonstrated by the registration. The points sampling considers both the local details and the overall shape of the model. The local details are captured by a graph-based segmentation, while for the overall shape, the approach voxelizes the model and samples points in terms of the entropy of the shape index of vertices in voxels. We present a number of results to show that the method significantly simplifies the surface without losing local details and global shape. The effectiveness of this method is finally demonstrated by the SoftICP and FICP registration results of overlapping range images.