MCL-3D: a database for stereoscopic image quality assessment using 2D-image-plus-depth source

摘要

A new stereoscopic image quality assessment database rendered using the2D-image-plus-depth source, called MCL-3D, is described and the performancebenchmarking of several known 2D and 3D image quality metrics using the MCL-3Ddatabase is presented in this work. Nine image-plus-depth sources are firstselected, and a depth image-based rendering (DIBR) technique is used to renderstereoscopic image pairs. Distortions applied to either the texture image orthe depth image before stereoscopic image rendering include: Gaussian blur,additive white noise, down-sampling blur, JPEG and JPEG-2000 (JP2K) compressionand transmission error. Furthermore, the distortion caused by imperfectrendering is also examined. The MCL-3D database contains 693 stereoscopic imagepairs, where one third of them are of resolution 1024x728 and two thirds are ofresolution 1920x1080. The pair-wise comparison was adopted in the subjectivetest for user friendliness, and the Mean Opinion Score (MOS) can be computedaccordingly. Finally, we evaluate the performance of several 2D and 3D imagequality metrics applied to MCL-3D. All texture images, depth images, renderedimage pairs in MCL-3D and their MOS values obtained in the subjective test areavailable to the public (http://mcl.usc.edu/mcl-3d-database/) for futureresearch and development.