Compression and Noise Removal Techniques for 3D Point Clouds

With the development of sensor technology for 3D imaging, the applications that involve use of these images have become a center of attraction for many developers. Current 2D image-based applications are being updated 3D imaging technology. With LiDAR’s and Dot Projectors paving their ways in smart phones, laptops and IoT devices, 3D imaging technology has become cheap and easily accessible. The importance of 3D imaging data is seen in medical applications, autonomous cars and robotics where precision is more important than efficiencies of the system. The precisions are obtained when a good amount of data is present of the surrounding and environment that the particular application is working with. The drawback of using 3D imaging data is that the current sensors collect numerous data which is difficult to process and consumes a lot of memory space. Processors which have the ability to process such data at high speeds are expensive and also have compatibility problems. Point Clouds and Meshes are the two widely used formats for applications based on 3D data. In this paper, we propose efficient methods that help us in reducing the noise, removal of unnecessary data and lossless compression of 3D image files that helps in efficient storage of files in the databases. The methods proposed in this paper also help in understanding the preprocessing steps for any recognition or classification applications that use Point Clouds or Meshes.