I am reading this book Discrete Cosine Transform [...]. K.R. Rao and P.YIP. 1990.

Given That

From Equation (2.4.11.a)

$$ X^{C(2)}(m) = \left ( \frac{2}{N} \right )^{1/2} k_m \sum^{N-1}_{n=0} x(n) \cos \left [ \frac{(2n+1)m \pi}{2N} \right ] $$

$m=0, ..., N-1$

Here, we have

$$ k_p = \begin{cases} \frac{1}{\sqrt{2}} & \text{when } p = 0 \text{ or } N \\ 1 & \text{otherwise.} \end{cases} $$

From Book's Equation (2.8.7)

$$ F^{C(2)}(k) = k_k \exp \left [ \frac{-j \pi k}{2N} \right ] \hat{F}_F(k) $$

Here, we have

$$ k_p = \begin{cases} \frac{1}{\sqrt{2}} & p = 0 \\ 1 & p = 1, ..., N-1 \\ 0 & \text{otherwise} \\ \end{cases} $$

Now Comes the following

It is not difficult to show that the sequence $f(n)$ can be recovered from the transformed sequence $F^{C(2)}(k)$, as as a result we have

From (Equation 2.8.8)

$$ f(n) = \sqrt{\frac{2}{n}} \Re \left \{ \sum^{2N-1}_{k=0} k_k F^{C(2)} \exp \left [ \frac{j \pi k}{2N} \right ] \exp \left [ \frac{2j \pi k n}{2N} \right ] \right \} $$

$n=0, ..., N-1$


  1. I do not understand how that $k_k$ is present on the (2.8.8). If I had to deduct this equation myself from (2.8.7) $k_k$ would pop on the other side of the equality inverted, as $k_k$, though it would yield a division by zero. But I could pretend (2.8.7) uses the same definition for as (2.4.11.a) resulting on a $F^{C(2)}(k)$ that is defined for $k > N-1$, then $k_k^{-1}$ wouldn't yield division by zero. How is (2.8.8) actually obtained?
  2. Why having two different definitions for $k_k$?
  • $\begingroup$ The two definitions of $k_k$ are equivalent in the range $0..N$ though (?) $\endgroup$
    – A_A
    Commented Jun 4, 2020 at 10:32
  • $\begingroup$ Indeed they are. However, I am kind of using a different version of the iDCT that uses the first definition, the second one wouldn't be suitable as I said in the question. Though I am not sure why the book does what it does. $\endgroup$ Commented Jun 4, 2020 at 16:44


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.