The Kinect has encountered widespread use in the areas of 3D mapping and computer vision thanks to its low cost, portability, and depth sensing capability. Unfortunately, its depth sensor has a maximum range of eight meters, and past four meters it tends to be very noisy. As a result, it performs poorly in open areas such as hallways where a large portion of the scene is too far away. This is not a problem if the sensor is being used for close-range collision avoidance, but it can be harmful when it is used to make larger-scale inferences about the surrounding area. The aim of this project is to use the information available in the depth and RGB images to estimate in real-time the missing information in the depth image, providing a more complete picture of what is present in a given scene

Hello! My name is Nishchal Bhandari. I chose my project based on previous UROP work I did in the SLI Lab at CSAIL. I am really excited to work on this project because the nature of it makes it useful for practical applications in mobile robotics. After I complete my project, I hope to have a greater knowledge and understanding of computer vision and image processing