The Use of Flowfield Motion Compensation for 3-D Stereoscopic Moving Image Compression

Using flowfield motion compensation, it is possible to efficiently code 3-D stereo pair moving images. The displacement fields between left and right eye images have a different character than frame-to-frame flowfield motion compensation, since motion blur, edge coverage, and other properties are inherently and fundamentally different. However, the general construction can be applied effectively using the flowfield coding method based upon a floating point codec (using windowed-sinc and wavelet hybrids, as described in the Jan/Feb 2009 SMPTE Journal). Both the inter-ocular displacement as well as the frame-to-frame motion can be built into a related flowfield set. For example, a motion compensation spine can be coded, with an inter-ocular displacement coded as a displacement to the left and right eye motion compensation fields. — Another useful approach is to code flowfields both forward and backward, as well as in relation to inter-ocular displacement fields. Such an approach requires resolving holes and confounds which appear during reverse flowfield traversal. However, this technique allows multi-reference motion compensation while needing to convey less flowfields. Additionally, flowfields can be delta coded, since they tend to have coherence between frames. Also, inter-ocular displacement fields are similarly coherent, and can similarly be delta coded. Further the flowfield and displacement flowfields and deltas are highly correlated. — The result is an efficient method of coding 3-D stereoscopic moving images.

Published

2009-10

Content type

Original Research

DOI

10.5594/M001349

ISBN

978-1-61482-943-0