A depth image quality benchmark of three popular low-cost depth cameras

Abstract

A depth camera outputs an image in which each pixel depicts the distance between the camera plane and the corresponding point on the image plane. Low-cost depth cameras are becoming commonplace and given their applications in the field of machine vision, one must carefully select the right device according to the environment in which the camera will be used given the accuracy these cameras can be associated with factors such as distance from the target, luminosity of the environment, etc. This paper aims to compare three depth cameras currently available in the market, Intel RealSense D435, which uses stereo vision to compute depth at pixels, ASUS Xtion and Microsoft Kinect 2 represent Time of flight-based depth cameras. The comparison will be based on how the cameras perform at different distances from a flat surface and we will check if the colour of the surface affects the depth image quality.