Abstract
Depth-Image-Based-Rendering (DIBR), as one important technique in 3D video system, can be used to generate virtual views. Unfortunately, the DIBR algorithms will introduce various distortions and induce an annoying viewing experience. And it has been proved that traditional 2D assessment quality metrics are not suitable for the DIBR-synthesized views. In this paper, we propose a novel approach to assess the quality of DIBR-synthesized images. The proposed method mainly considers three kinds of DIBR-related distortions: holes distortion, strip-sharped distortion and global sharpness. Holes and strip distortions as two local features are used to characterize the local quality of DIBR-synthesized image, respectively. For the global sharpness we consider the Just Notice Difference (JND) model of human eyes and use it to extract the JND-based global difference for analyzing the global quality. Finally, we combine the holes distortion evaluation, strip distortion evaluation and global quality to infer the overall perceptual quality. Extensive experiments indicate that our method achieves higher accuracy of quality prediction than most competing metrics.
Highlights
Three dimensional (3D) video system provides viewers with more immersive experience of natural 3D scene, so it has received considerable attention in recent decades [1], [2]
IRCCyN/IVC database is the first released benchmark database for DIBR synthesized image evaluation. It provides 12 original images selected from three MVD sequences (‘‘Book Arrival’’, ‘‘Lovebird1’’ and ‘‘Newspaper’’) and 84 DIBR-synthesized images generated by seven DIBR algorithms
IETR database contains 10 original images and 140 synthesized images generated by eight DIBR image synthesis algorithms, including Criminisi’s [32], LDI [33], Ahn’s [34], Luo’s [35], HHF [36], VSRS [37], Zhu’s [38] and VSRS2 [37]
Summary
Three dimensional (3D) video system provides viewers with more immersive experience of natural 3D scene, so it has received considerable attention in recent decades [1], [2]. L. Wang et al.: Quality Assessment for DIBR-Synthesized Images With Local and Global Distortions structural information was preserved in the retargeted image. Li et al [14] through measuring the Local Geometric distortions in disoccluded regions and Global Sharpness (LOGS) proposed a new quality model for DIBR-synthesized view images. Yue et al [17] analyzed the categories of DIBR-related distortions and proposed a No-Reference (NR) quality evaluation method for the DIBR-synthesized images. With the extracted distortion regions, we proposed a quality evaluation model of DIBR-synthesized images.
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
