Digital BW, The Print

Mark Savoia wrote:


> Thanks for that reminder, I don't think I could take another month of  
> that. :)

Then why ask essentially the same question again?

The answer of course is "it depends." One algorithm is going to have an 
advantage over the other when run against a dataset that has certain 
characteristics. Genuine Fractals seems (to me, on average) to do best 
on images with a lot of high frequency information; conversely, bicubic 
algorithms tend to alias more easily on such inputs even when they are 
smoothed, and when smoothed seem to have trouble preserving the highest 
frequency information present in a set of data. OTOH, bicubics tend to 
process gradients and other low frequency information with more 
reliability than I'm used to seeing from GF.

Neither algorithm is considered to be very ideal for imagery at this 
point, and both are widely considered to be suboptimal for most imaging 
applications. If you have both, why don't you try them both against your 
image of interest and report back to us? If you lack one or both, just 
don't go there unless you feel a need to live in the middle ages; get 
yourself a program that supports lanczos and s-spline interpolation.

-- 
Jeff Medkeff
Eagle River, Alaska