Video Signal Processing Inter-View Prediction Depth Data

Instead of extracting disparity vectors from the bitstream and using them for prediction, the patent inverts the approach by deriving disparity vectors from depth data. This reversal allows the system to achieve accurate inter-view prediction while reducing the bit requirements for coding disparity vectors, as the depth data serves as the primary source rather than the disparity vector.

The patent changes the fundamental parameter used for inter-view prediction from disparity vectors to depth data. By using depth data as the basis for prediction and only coding the residual difference, the system achieves both high prediction accuracy and reduced bit requirements, resolving the contradiction between these two parameters.

The patent applies different coding precision to different regions of the depth data based on their importance. Motion regions are identified and handled differently from non-motion regions, with motion regions receiving higher precision coding where needed and lower precision coding where depth data can be derived from neighbor views, thus reducing overall coding complexity while maintaining necessary accuracy.

Instead of coding all depth data with maximum precision, the patent applies partial precision coding by deriving depth data for non-motion regions from neighbor views and only coding the residual differences for motion regions. This partial action approach reduces coding complexity and transmission requirements while maintaining sufficient accuracy for the critical motion regions.