encoding and evaluation approach

Currently we examined only the publicly available (uncompressed) YUV test sequences with the TML-97 encoder release. Since these sequences are of short length, the statistical evaluations done are to be concerned of being only early-stage evaluations. The same is true regarding the encoder itself since it is not the final ITU-T approven encoder due to the fact that the H26L standard is due later in 2002. Nevertheless these early evaluations will provide us with a testbed for more thorough evaluations once the standard is final. We emphasize that we did not encode in real-time; thus there was no encoder bottleneck. For each video we encoded the YUV information into an H26L bit stream with the TML-97 video software; we encoded each video at different quality levels (quantisation parameters). These quantisation parameters were equally set for all different frame types (I,P,B) at 7 different levels, ranging from 1 to 31. We used a GoP-structure utilized in MPEG4 video-encoding for future comparison (IBBPBBPBBPBB). The encoder output was used to gather the information needed to perform the statistical evaluation, working on a parsing software has been postponed until the standard is final.