Reply by Christian Gollwitzer October 7, 20122012-10-07
Am 07.10.12 07:17, schrieb manishp:
> Sirs, > > Sometime back there was a dicussion about most of the filtering operating > in time domain. > > Now, in case of JPEG compression, the time domain samples are first > converted into frequency domain and then processed further for > compression. > Why is it not possible to apply direct filtering on the time domain > samples?
It is, but it would be more complicated, and after all, the samples in frequency domain are stored. The whole point of JPEG compression is that in frequency domain, it is easier to decide which information to store and which to throw away. Note that the quantization is non-linear filtering that cannot be expressed in terms of the ususal FIR, IIR concepts, which are all based on liner response; therefore our beloved tools like impulse response, Green's functions etc. do not work. Christian
Reply by manishp October 7, 20122012-10-07
Sirs,

Sometime back there was a dicussion about most of the filtering operating
in time domain. 

Now, in case of JPEG compression, the time domain samples are first
converted into frequency domain and then processed further for
compression.
Why is it not possible to apply direct filtering on the time domain
samples?

Thanks, Manish