replace zeroes with red noise before 2D FFT?

I came across a claim in another forum for dealing with 2D FFT where there
are dead traces (data is t-x domain before FFT). So if there is a dead
channel, fill it with red noise whose amplitude matches roughly the other
channels, then do the 2D FFT. 
Supposedly this minimises artefacts.
Plausible?

 
>I came across a claim in another forum for dealing with 2D FFT where there
>are dead traces (data is t-x domain before FFT). So if there is a dead
>channel, fill it with red noise whose amplitude matches roughly the other
>channels, then do the 2D FFT. 
>Supposedly this minimises artefacts.
>Plausible?

What then happens with the FFT result?  Is this part of compression /
decompressions?

Steve