On Tuesday, November 4, 2003 at 1:29:40 PM UTC-6, Matthias Alles wrote:> Hi! >=20 > Can anyone explain me, why you use DCT instead of DFT for transformation==20> of image-data (like it is done for JPEG). I know that the DCT can be=20 > performed by a DFT with a post-rotation. But what are the advantages of==20> the DCT compared to the DFT? Are there any benefits from the post-rotatio=n?>=20 > Regards, > MatthiasJPEG isn't really using the DCT in its full generality but rather a highly = specialized 8x8 table which can just as well be expressed as without the ne= ed or use of trig (and tabulated efficiently without trig). It's just a way= to get LOCALIZED transforms, but without employing the full machinery of t= ime-frequency analysis (or the 2D equivalent of spectrographs). The 8x8 table is simply this: DDDDDDDD GECAaceg FBbffbBF EagcCGAe DddDDddD CgAEeaGc BfFbbFbB AcEgGeCa where -a =3D A =3D 1/4 root(2 - root(2 + root(2))) -b =3D B =3D 1/4 root(2 - root(2)) -c =3D C =3D 1/4 root(2 - root(2 - root(2))) -d =3D D =3D 1/4 root(2) -e =3D E =3D 1/4 root(2 + root(2 - root(2))) -f =3D F =3D 1/4 root(2 + root(2)) -g =3D G =3D 1/4 root(2 + root(2 + root(2))). The coefficients satisfy simple algebraic relations with one another which = allows for efficient methods that factor the matrix or use other tricks sim= ilar to those used for the Fourier Transform.
DCT vs. DFT
Started by ●November 4, 2003
Reply by ●June 22, 20162016-06-22