User guide

MACROBLOCKS AND MOTION

The encoder divides a frame into a grid of blocks, called macroblocks,

and attempts to track the motion of each macroblock in the current

frame back to a matching area of picture in the previous frame. This

process is known as the motion search.

When the encoder finds a matching area for a block in the previous

frame its position is recorded by use of a vector. A vector is simply a

numerical representation of direction and magnitude. For example

the vector (4, -7) might mean right 4 units, down 7 units.

Because these vectors happen to represent the movement of a

macroblock they are called motion vectors. In the illustrations the

red arrows represent the motion vectors.

Storing the motion vector of each block from frame-to-frame allows

DivX to recreate much of each new frame from the last using far less

data than if each block were stored as a complete image. The process

is similar to cutting up one frame into little pieces, then maneuvering

them like a make-shift jigsaw puzzle using motion vectors in order

to recreate the next frame. Of course the resulting picture is not

perfect and DivX has to store some additional data to correct for

the difference (known as the residual). Even so, recreating the

picture using motion vectors and residual consumes far less

storage than recording the entire image for every single frame.

DivX 10

DivX, LLC User Guide