Description
With the rapid growth of wireless multimedia services, the demand for video transmission over wireless channels has been increasingly exponentially. This has led to the design of error resiliency schemes for obtaining better codecs. The perceptual video quality is influenced not only by the compression artifacts, but also by the channel errors. Hence, a video codec needs to accommodate contradictory requirements: coding efficiency and robustness to data loss, along with other limitations such as memory, bandwidth, real time coding and complexity. In this thesis, we have studied the comparative performance of two prominent open source H.264 AVC video codecs - Joint Model (JM) and Fast Forward Motion Picture Expert Group (FFMPEG). We have studied various features of these codecs, such as macroblock tree rate control, multi-pass encoding, and psychovisual optimization model (it includes the parameters like adaptive quantization, psy-RD and psy-trellis). Observations based on the PSNR values and rate-control have been used to analyze the objective video quality for error-free and corrupted bitstreams generated by both codecs. In addition, we have analyzed the creation of different priorities based on cumulative mean squared error (CMSE) values for dividing the I, P and B video frames into four priorities. These studies helped us in understanding which codec can be more robust against channel errors or losses and how we can improve their performance.
