Operation
Global motion compensation describes the motion in a scene based on a single affine transform instruction. The reference frame is panned, rotated and zoomed in accordance to GMC warp points to create a prediction of how the following frame will look. Since this operation works on individual pixels (rather than blocks), it is capable of creating predictions that are not possible using block-based approaches.
Each macroblock in such a frame can be compensated using global motion (no further motion information is then signalled) or, alternatively, local motion (as if GMC were off). This choice, while costing an additional bit per macroblock, can improve prediction quality and therefore reduce residual.
Because the transforms used in global motion compensation are only added to the encoding stream when used, they do not have a constant bitrate overhead. A predicted frame which uses GMC is called an S-frame (sprite frame) while a predicted frame encoded without GMC is called either a P-frame, if it was predicted purely by previous (past) frames, or a B-frame if it was predicted jointly with past and future frames (an unpredicted frame encoded as a whole image is referred to as an I-frame).
Read more about this topic: Global Motion Compensation
Famous quotes containing the word operation:
“You may read any quantity of books, and you may almost as ignorant as you were at starting, if you don’t have, at the back of your minds, the change for words in definite images which can only be acquired through the operation of your observing faculties on the phenomena of nature.”
—Thomas Henry Huxley (1825–95)