Authors: NÜKHET ÖZBEK, GİZEM ERTAN, OKTAY KARAKUŞ
Abstract: In 3D perception, the binocular suppression of the human vision, perceiving high quality 3D video in the case of a view in higher quality, can be exploited in asymmetric stereo video coding for efficient 3D rate scaling. Hence, the best stereo rate-visual distortion performance may be gained by asymmetric coding, which is the reduction of the spatial and/or quantization resolution of the low-quality view, while keeping the high-quality view in full resolution. However, how to determine the level of asymmetry and what type of scaling should be chosen are still in question. In this work, we try to assess the overall performance of the scalability options with several test contents. The test videos are encoded at critical bitrates with symmetric options and spatial or signal-to-noise ratio (SNR) asymmetric coding, and are subjectively evaluated in a stereo-polarized projection 3D display system. Two different types of evaluation methodologies are used: the Double-Stimulus Continuous-Quality Scale (DSCQS) and Subjective Evaluation of Stereo Video Quality (SESVIQ). Dense visual tests show that the spatial scaling is generally inferior when compared to SNR scaling, except that high motion scenes and symmetric SNR are more preferable for a higher bitrate. The characteristics of the video content should be taken into consideration for efficient stereo rate scaling.
Keywords: Asymmetric stereo video coding, 3D video quality assessment, DSCQS, SESVIQ, blur, blockiness
Full Text: PDF