Some questions about the ica method

[WayneSzzz]

for c=1:3
RGBDetrend= spdetrend(RGB(:,c),Lambda); %M. P. Tarvainen, TBME, 2002
RGBNorm(:,c) = (RGBDetrend - mean(RGBDetrend))/std(RGBDetrend); %normalize to zero mean and unit variance
end

Because the average pixels of each frame of RGB channels are very similar，The variance is close to infinity, and ica calculation cannot be performed. How do you deal with it?
I Use UBFC too.