An algorithm is presented for switching of compressed video streams in order to support the insertion of local programs and commercials for digital television production. This enables the creation of particular points in compressed video streams, whose characteristics allow to seamlessly switch between streams. These switching points are constructed by computing the buffer bounds in the video encoder so as to ensure that the decoder buffer does not underflow or overflow.

We investigate the problem of MPEG-2 video quality degradation due to re-encoding using different coding parameters. We examined the statistical properties of pre-coded video and discovered that the coding parameters can be recovered from pre-coded video because the compressed video is represented by a limited set of parameters. We present an algorithm to detect if a picture is intra-coded or original and extract coding parameters if the picture is intra-coded. We show with numerical results that the coding parameters can be effectively extracted from pre-coded video and used to reduce additional coding loss, thus greatly improving the performance of multiple-generation coding.

We propose a new approach for creating splice points in compressed video streams. The resulting streams can be switched seamlessly, and therefore no decoder underflow or overflow events occur. The proposed variable splicing parameters (VSP) method eliminates major encoder buffer constraints and related video encoding issues associated with the standard approach for creating seamless splice points. These advantages are achieved at the expense of a minor increase in the splicing operation complexity. When using VSP splice points, the splicer has to wait a specific period of time (equal to the difference between the out point and in point decoding delays) before switching into the new stream (i.e. the stream with the in point).

An algorithm is presented for increasing the overall video quality of multiple variable rate video encoders multiplexed onto a single constant bit-rate channel. This enables an encoder, that is processing complex video, to use more of the channel bandwidth than an encoder that is compressing less complex video. The encoder rates are continually adapted. A rate controller determines the channel rate for each encoder as well as the target rate allocated to individual pictures encoded by the encoders. The target rates for individual pictures are computed in order to maintain constant quality across encoders while the encoder channel rates are derived so as to avoid encoder and decoder buffer underflow and overflow.

In general, growth algorithms for optimal tree-structured vector quantizers do not exist. In this paper we show that if the source satisfies certain conditions; namely, that of diminishing marginal returns; optimal growth algorithms do exist. We present such an algorithm and compare its performance with that of other tree growth algorithms. Even for sources that do not meet the necessary conditions for the growth algorithm to be optimal, such as for speech with unknown statistics, it is seen by simulation that the algorithm outperforms other known growth algorithms, For sources that do not satisfy the required conditions, the algorithm presented here can also be used to grow the initial tree for the pruning process. The performance of such pruned trees is superior to that of trees pruned from full trees of the same rate.>

An image coding system is proposed in which hexagonally sampled images are decomposed into subbands that are selective in both frequency and orientation. Nine high-frequency bands each of one octave and 60 degrees of angular orientation are built on a pyramidal structure along with one low-frequency band. Bits are then allocated optimally to the subbands under a mean-squared-error distortion measure. Optimal quantizers are designed for each subband for entropy coding under a Laplacian distribution assumption. An adaptive version of the coding scheme is also implemented, and the results are compared at rates of 0.5 and 1.0 bits per pixel.>