70.2K

Verified Solution

Link Copied!

Q3) [15 pts] Assume that you have a signal s = (s(1), (2), ....s(n)). Let us represent its DCT transform ast=(t(1),t(2), ....t(n)). Suppose you remove 25% of coefficients in t, set them to O's. The result is tt =(t(1),t(2), .t(k), 0, 0, 0, ....0) where k= 0.75*n. Use this information to answer the following Questions (A and B). A) [3 pts] Let (dl) be the distortion in transform domain. Write down the expression for (dl) in terms of t's. B) [3 pts] Let us reconstruct the signal by applying the inverse transform on tt. Let us say, the output is ss. Suppose you compute the mean square error (d2) between s and ss. Is d2d1? Why? C) [3 pts] Is DCT better than KLT? Why? D) [3 pts] Is DWT better than DCT? Why? Q3) [15 pts] Assume that you have a signal s = (s(1), (2), ....s(n)). Let us represent its DCT transform ast=(t(1),t(2), ....t(n)). Suppose you remove 25% of coefficients in t, set them to O's. The result is tt =(t(1),t(2), .t(k), 0, 0, 0, ....0) where k= 0.75*n. Use this information to answer the following Questions (A and B). A) [3 pts] Let (dl) be the distortion in transform domain. Write down the expression for (dl) in terms of t's. B) [3 pts] Let us reconstruct the signal by applying the inverse transform on tt. Let us say, the output is ss. Suppose you compute the mean square error (d2) between s and ss. Is d2d1? Why? C) [3 pts] Is DCT better than KLT? Why? D) [3 pts] Is DWT better than DCT? Why

Answer & Explanation Solved by verified expert

See Answer