Bilinear or bicubic?

I'm using the Minolta Dimage Viewer software to resize images; some

down, for Photo.net postings, and some up, for enlarged printing.

When is it better to use bilinear mode, and when bicubic mode?

Minolta's software gives me the choice. I'm somewhat confused,

after searching through a few posts on this subject that went into

other areas of discussion, i.e. Genuine Fractals, etc.

bi-cubic is always better.

Bicubic smoother for sizing up, Bicubic sharper for sizing down. I found GF Pro to be

basically worthless.

Vuescan's downsample technique seems to be similar to Photoshop's "nearest neighbour", a distance third in the smoothness race, as noted here:

 

http://www.photo.net/bboard/q-and-a-fetch-msg?msg_id=00BcsU

...and all of 'em are pretty outdated. These software companies should read some scientific papers: http://graphics.cs.msu.su/en/research/scaling/index.html

Take a lood at this.

 

ftp://ftp.bmtmicro.com/bmtmicro/DLC_WEB/Picture%20Window%20(BMT%20Micro)/Picture%20Window%20Doc/Resampling.pdf

For certain resizing values, e.g. (depending on software) 25% 33%

50% 67% 75% and 200%, bilinear and bicubic produce identical results.

Because bilinear is faster, you might as well use that. Emre has a

good point. At values not mentioned above, Lanczos seems better for

downsampling than Bicubic Sharper, and certain extra-cost algorithms

beat Bicubic Smoother for upsampling.

<p>Bilinear works well with schematic, bicubic is the one to use for photos.

<p>--ben

Thanks, everyone. I followed your links, where suggested.

'nother link:

 

http://www.xs4all.nl/~bvdwolf/main/foto/down_sample/example1.htm