JVET-C0062 Improved affine motion prediction [F. Zou, J. Chen, M. Karczewicz, X. Li, H.-C. Chuang, W.-J. Chien (Qualcomm)]

This document presents an improved affine motion prediction scheme in the current reference software JEM2.0. A 6-parameter affine motion model is used in addition to the existing 4-parameter affine motion model. The scheme was tested on top of JEM2.0. Simulation results reportedly show that the proposed scheme achieves 0.3% BD rate reduction on the random access configuration in JEM2.0, and 0.5% BD rate reduction on the random access configuration in an HM test.

Encoder runtime increased by approximately 10%

This proposal switched between 4 and 6 parameter models at the PU level, only for the 2Nx2N case. In the case of QTBT, it would always be applied. In the case of 6 parameters, three corner vectors have to be signalled.

It was reported that always using 6 parameters does not have good performance.

In affine merge mode, no additional signalling is necessary.

The cross-checker reports that in affine merge mode, the order is modified.

Higher gain was shown for class A2.

Investigation in an EE was encouraged, potentially including additional material with affine motion for better understanding of the effects.