Reduced-reference Based Video Quality-metrics Using Representative-luminance Values

This paper proposes an objective video quality-metrics method based on a reduced-reference model for measuring end-user video quality in video applications such as IPTV. It chooses a representative-luminance value for each original video frame and transmits that value, along with information regarding the positions of all pixels having that value, to end-user terminals. End-user PSNR is estimated on the basis of that transmitted information. Since the space domain frequency of the luminance correlates with quantization errors generated in the encoding process, quantization errors in medium frequency regions are close to an average quantization error for an entire frame. Therefore, representativeluminance values are chosen from among medium frequency blocks to achieve accurate estimation for average errors. Simulation results show average PSNR estimation error to be -0.12 dB and the standard deviation of estimation error to be 0.97. This indicates that accurate PSNR estimation can be achieved with the proposed method.