Complexity-quality analysis of transcoding architectures for reduced spatial resolution

This paper presents a detailed complexity-quality analysis of various video transcoding architectures that perform spatial resolution reduction. In addition to a reference architecture, two additional architectures are considered. The reference architecture is a cascaded approach that decodes, down-samples and re-encodes the video. The first additional proposal, intra refresh architecture, compensates for various errors by converting select macroblocks to intra-coded blocks. The second proposal, partial encoder architecture, is similar to the reference, but simplifies the re-encoding process by not compensating for re-quantization errors. The optimization of several processing components, including the down-sampling operation, is presented. We have found that the two additional architectures provide a good trade-off in terms of complexity and quality compared to the reference.