Reduced-reference quality evaluation for compressed depth maps associated with colour plus depth 3D video

3D video quality assessment is a key task in transmission of immersive video, both for the assessment of end-to-end system performance and for quality evaluation “on the fly”. For example, measured 3D video quality at the receiver side can be used as feedback information for controlling transmission parameters. However, measuring video quality at the receiver side using Full-Reference quality metrics (e.g. PSNR) is less practical due to the need of original image sequence at the receiver. Hence, Reduced-Reference and No-Reference quality metrics are being developed to evaluate the quality. Since edges and contours of the depth map represent different depth levels, this paper proposes a Reduced-Reference quality metric for the depth maps associated with colour plus depth 3D video using edge detection. Edge/Side information (i.e. binary edge mask) generated using sobel filtering is utilized. Results show that the proposed metric can be used in place of a Full-Reference metric with an acceptable accuracy.