What is the advantage of applying 1 DCT for rows & columns?

[amith]

Hi all,

what is the advantage of applying 1 dct for rows & colums seperatly , why cant we take a 2-dct directly , if we have nxn matrix what will be the complexity
of applying 2d-dct directly when campared to seperate rows & columns .

 

[Sathish]

Re: DCT

hi,
The explanation is very simple,

2D:
(u,v) = Σ Σ f(m,n) cos((2m+1)uΠ/2N) cos((2n+1)vΠ/2N), consider it is 8x8 block, then , for computing one output value the expression has to do 8x8 = 64 manupulations and to complete one whole block i.e. 8x8 it has to do 64 * 64 manupulations that is 4096.

1D:
in case of 1D one output value requies 8x8(row wise) + 8x8(column wise) manupulations i.e. 128, hence the total manupulation is 128 * 8 i.e. 1024.


this is the computational complexity you asked

 

[Ralph80]

Re: DCT

See this DCT overview

 

[phongphanp]

DCT

Hi,

The 1D processing is slow than block processing for huge data , O.K. It's easy and be main design for sound processing.

The block processing is good for image processing , and other jobs .