[LS2N_IPI_DIBR_Videos] 3D-TV System with Depth-Image-Based Rendering Architectures, Techniques and Challenges

Presentation & Goal The IRCCyN IVC DIBR Videos quality database contains 102 video sequences of 6 seconds in 1024x768 resolution with between 15 and 30 frames per second. Individual votes and Mean Opinion Scores (MOS) obtained by an Absolute Category Rating with Hidden Reference (ACR-HR) experiment are provided. Three different multiview plus depth (MVD) sequences are considered in this database. The sequences are Book Arrival (1024x768, 16 cameras with 6.5cm spacing), Lovebird1 (1024x768, 12 cameras with 3.5 cm spacing) and Newspaper (1024x768, 9 cameras with 5 cm spacing). Seven DIBR algorithms processed the three sequences to generate, for each sequence, four new viewpoints. Initial goal This database was done to consider the reliability of usual assessment methods when evaluating virtual synthesized views in the multiview video context. Virtual views are generated from Depth Image Based Rendering (DIBR) algorithms. Because DIBR algorithms involve geometric transformations, new types of artifacts come up. The question regards the ability of commonly used methods to deal with such artifacts. Possible usage This video quality database can be used to test a video quality metric precision in a context of DIBR. Dataset description We used Absolute Category Rating - Hidden Reference (ACR-HR) and as test methodology. We provide a spreadsheet with the individual scores, the Mean Opinion Score (MOS) and the Differential Mean Opinion Score (DMOS) for each of the 102 Processed Video Sequences (PVS). The PVS (avi videos) are available on our FTP server and are in the AVI format. The information about the source contents, video tested generation, subjective experiment environment and available data are provided in a description file. Source contents (SRC) Source number Source name Description 01 Book Arrival Description : A man in an office standing up to welcome another man. Author : Heinrich-Hertz-Institut. Video frame rate : 15Hz Number of frames : 100 02 Lovebird1 Description : A couple walking in front of a temple. Author : Electronics and Telecommunications Research Institute (ETRI). Video frame rate : 30Hz Number of frames : 150 03 Newspaper Description : A man and a girl reading the newspaper around a table when a man coming behind them. Author : Gwangju Institute of Science and Technology (GIST). Video frame rate : 30Hz Number of frames : 200 Hypothetical Reference Circuits (HRC) The Processed Video Sequences (PVS) were created using the following Hypothetical Reference Circuits (HRC). HRC name HRC description cam_xx [No impairment - Reference sequence view 1] cam_yy [No impairment - Reference sequence view 2] cam_zz [No impairment - Reference sequence view 3] A1 Fehn cropped : where the depth map is pre-processed by a low-pass filter. Borders are cropped, and then an interpolation is processed to reach the original size. A2 Fehn interpolated : Borders are inpainted by the method proposed by Telea A3 Tanimoto et al., it is the recently adopted reference software for the experiments in the 3D Video group of MPEG. A4 Müller et al., proposed a hole filling method aided by depth information. A5 Ndjiki-Nya et al., the hole filling method is a patch-based texture synthesis. A6 Köppel et al., uses depth temporal information to improve the synthesis in the disoccluded areas. A7 corresponds to the unfilled sequences (i.e. with holes). cam_xx_qp26 Reference sequence view 1 coded by H.264 with a QP = 26 cam_xx_qp34 Reference sequence view 1 coded by H.264 with a QP = 34 cam_xx_qp44 Reference sequence view 1 coded by H.264 with a QP = 44 There are several references (3 first lines of the table) for each content, because there are the 3 views used for each content. So there are 3 HRC with no impairment. Experiment information Features Value Highlights of this database Depth Images Based Rendering Free-viewpoint video quality Video Frame frequency 30 or 15Hz Video format 1024x768 Video length 100, 150 or 200 frames Video duration 5 to 7 seconds Number of videos 102 Subjective test Observation distance 4H Environment ITU-R BT.500-11 Background luminance 25cd/m² Methodology ACR Duration 30 minutes Pre screening Snellen, Ishihara Post screening (Rejection method) VQEG Multimedia Observers Number of observers 32 naives Display Display TVLogic LMV401 Display resolution Full HD (1920x1080 pixels) Display frequency 60Hz Display mode progressive Display luminance max 180cd/m²

搜集汇总

背景与挑战

背景概述

该数据集是一个包含102个视频序列的质量评估数据库，专门用于测试深度图像渲染（DIBR）算法在3D多视角视频中的视频质量。数据集基于三个多视角加深度序列，通过七种DIBR算法生成新视点，并提供主观实验的评分数据，旨在评估虚拟合成视图质量评估方法的可靠性。