Question

$263516365_4538440286241711_8713661674698574175_n.jpg$ PROBLEM 4.1: (3 points) Consider the application of DPCM to an image as suggested by the figure shown below.

We will base our prediction of pixel A on pixels B and D. Assume that the correlation between pixels is known and given by ρAB = ρAD = ρ ρBD = ρ 0 Following our discussion in class, what are the optimal fixed predictors (B, D) for this 2-D DPCM system

PROBLEM 4.2: (3 points) Consider the following 8 × 8 image x(n1, n2) corresponding to the (n − 1) frame of a video sequence: x(n1, n2) = ( 2, 2 ≤ n1, n2 ≤ 5 1, otherwise Consider also the following associated motion vectors corresponding to each 2×2 block (i, j), where (i, j) = (0, 0) corresponds to top left block and (i, j) = (3,3) corresponds to bottom right block: MV (i, j) = ( [1, 1], 1 ≤ i, j ≤ 2 [0, 0], otherwise (a) (2 points) Using the motion vectors above, construct the motion compensated image for frame n. (b) (1 point) Consider a video sequence with frame size 512 × 512 pixels. If motion blocks of 16 × 16 are used, how many motion vectors must be transmitted as side information?

(3 points) Let x(n1, n2) represent an 8×8 luminance block. The DCT of x(n1, n2) is quantized using the JPEG luminance quantization Q matrix as specified in Section K, Table K.1, of the ISO/ITU-T JPEG standard document that is provided on the course web site. The resulting quantized DCT of x(n1, n2) is given by: ...