Generation of synthetic Kinect depth images based on empirical noise model. Issue 13 (1st June 2017)

• Record Type: Journal Article
• Authors: Iversen, T.M.; Kraft, D.
• Is Part Of: Electronics letters. Volume 53:Issue 13(2017)
• Page Start: 856
• Page End: 858
• Publication Date: 2017-06-01
• Subjects: computer vision -- image sensors; Kinect v1 depth images -- computer vision algorithms -- empirical noise model -- synthetic Kinect depth images; Electronics -- Periodicals; 621.381
• DOI: 10.1049/el.2017.0392
• Languages: English
• ISSNs: 0013-5194

Abstract

Abstract : The development, training and evaluation of computer vision algorithms rely on the availability of a large number of images. The acquisition of these images can be time‐consuming if they are recorded using real sensors. An alternative is to rely on synthetic images which can be rapidly generated. This Letter describes a novel method for the simulation of Kinect v1 depth images. The method is based on an existing empirical noise model from the literature. The authors show that their relatively simple method is able to provide depth images which have a high similarity with real depth images.