This paper investigates compression of depth images -particularly, noisy depth images -captured by depth sensors like Kinect. Our scheme is based on incrementally detecting planes in depth images and then storing the plane parameters instead of the depth information of the pixels lying on the detected planes. Residuals are then computed and compressed using standard image compression techniques. Our technique incorporates the input error model for comprehensive and accurate plane detection. Thereby, this accounts for the reliability of the input data in the compression scheme. The plane detection also accounts for edges. Experiments exhibit better image quality than standard compression techniques with smaller error. We additionally propose a novel error metric to evaluate compression of noisy depth images.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.