MPEG interlaced video transcoding for a networked video browsing system
Atsushi Shimizu · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2004
This paper proposes a transcoding technique with low computational complexity, by means of eliminating high order DCT coefficients. Although the computational complexity of the DCT coefficient reduction technique is very low, this technique induces severe drift errors and produces poor-quality blockish pictures; artifacts are especially noticeable in interlaced video. Therefore, we propose a fast transcoding method of interlaced video which can reduce these artifacts. Our technique detects problem blocks, and allocate them with higher order DCT coefficients into the new MPEG stream. As a result of using our proposed technique, experiments show that artifacts caused by drift errors are mostly reduced, and the computational complexity is very low. Only 4.1% of all the luminance blocks are detected as problem blocks. Hence, most blocks are filled with the minimum set of DCT information required in the transcoded stream. Moreover, we develop a networked video browsing system by using our proposed transcoding method to play multiple streams in a client viewer, simultaneously.