A fast encoder of frame-compatible format based on content similarity for 3D distribution

• A fast encoding algorithm is proposed for frame compatible format (FCF).
• Inter-view prediction correlations are exploited to reduce the encoding complexity.
• It is independent from depth maps and camera parameters to facilitate FCF coding.
• The algorithm is stable to videos captured by either moving cameras or static cameras.
• It provides 72.45% and 81.08% reduction in Intra and Inter coding complexity.

Frame compatible format (FCF), which multiplexes the left and right eye views into one frame spatially, has been considered as one of the most promising solutions for 3D video broadcasting. In this paper, a fast algorithm exploiting the content similarity is proposed to reduce the computational complexity for coding FCF. A low complexity average shift estimation method is proposed to find the matching blocks in the two interleaved views, based on which mode candidate sets are designed for accelerating intra and inter coding of the second interleaved view (the interleaved view at the bottom/right in FCF) by exploiting inter-view correlations. The proposed algorithm achieves an average of 72.45% reduction in Intra coding complexity and an average of 81.08% reduction in Inter coding complexity for the second interleaved view, respectively, with negligible loss in compression efficiency. By combining with conventional fast algorithms, that is applying an conventional fast algorithm to code the view on the top/left and the proposed one to code the view on the bottom/right, the whole encoder provides up to 5.44 times higher encoding speed relative to applying the conventional algorithm solely to the whole frame with negligible performance degradation.