1.In most video sequences there is a little change in the contents of image from one frame to the next.
2.Most video compression schemes take advantage of this redundancy by using the previous frame to generate a prediction of current frame.
3 Consider two frames from a motion video sequences as shown below:
$$\text{Figure 3.1.a}$$
4.The only difference between the two frames that the face has moved slightly downwards to right of the frame, while the triangular object has moved to left.
5.If we simply take the difference between the frames, the displacement of object from the frame result in an image that contains more detail than the original image.
$$\text{Figure 3.1.b}$$
6.Hence differencing equation instead of reducing information, there is actually more information that needs to be transmitted.
7.Thus in order to predict the pixel values in current frame based on the previous frame ‘Block Based Motion Compensation’ is used.
8.In this approach, the frame is being encoded is divided into blocks of size M × M.
9.For each block we search the previous reconstructed frame that closely matches the current
frame.
10.We can measure distance between two blocks by the absolute difference corresponding to the pixels of 2 blocks.
11.If distance between d (current block, previous block) > > (prespecified threshold, the block) is declared uncompensable and is encoded without prediction. This information is transmitted to the receiver.
$$\text{Figure 3.1.c}$$
12.The encoder scans the current frame block by block.
13.For each block B, it searches the preceding frame for identical block C.
14.For such block, encoder writes the difference between its past and present location as:
$((C_x – B_X , C_y – B_y) = ( Δ_x, Δ_y)$
Where $( Δ_x, Δ_y)$ is the motion vector.
15.Motion Compensation is effective if objects are just translated , not scaled or related.
Motion compensation is lossy.
16.E.g (15, -4) means a particular block has moved from reference pictures to current picture by displacing it is pixels to current by displacing if is pixels to right and 4 digits up (otherway sound as of Cartesian around system)
Figure 3.1.d Motion Compression With Respect to Image Compression
and 3 others joined a min ago.
teamques10
★ 68k