S.S. Zwezhinskii, Yu.A. Ivanov, M.A. Smichek, A.N. Rudnev

In this paper the influence of packet errors on visual quality of Н.264 video is studied and the analytical model explaining this effect is developed. Single packet loss as well as single-bit errors are not able to adequately simulate the channel affected by fadings. As a rule, errors often are long-time as the high probability of packets loss happens in specific phase of transmission, for example at the bad propagation. Damping of a transmittable signal leads to grouping of errors. Group of error packets is the sequence of the packets not received or received with an error by transmission on communication path in a certain time interval. In this regard, the concept of errors group length BEL is introduced, which is defined as number from the first to the last error packet inclusively in certain group of errors. Research of influence of a packet error on the received quality was carried out by simulation modeling of transmission of 30-minute video through a wireless network with casual burst errors in the channel. Packet error simulation was performed by manually deleting UDP packets from the reception data file, allowing to explore and analyze the visual quality change when packet loss. At the same time received and undistorted file can be obtained in the transmission on the "ideal" channel with unlimited bandwidth and no delay by means of NS-2 software, and then manually remove packets according to certain scenario. Quality estimation was carried out by means of indicators PSNR and MOS, calculated by using the hardware and software tools. It is shown that at an estimation of influence of erratic packets on the received quality it is necessary to analyze not only probability of appearance of errors, but also their structure and length of grouping. It is shown that influence of group of errors on average quality stronger, than at single errors at the expense of local errors concentration. Average quality of video sequence thus increases in process of magnification of grouping length at invariable value of burst errors appearance probability. At grouping length BEL ≥ 60 average quality is almost identical to original video. The magnification of index BEL leads to magnification of video sequence average quality irrespective of an index of burst errors probability. It is shown that for an estimation of video quality in the conditions of errors batching in actual practice it is necessary to research real allocation of errors batching in communication path.