What's the differences between the scaling algorithms?

Here everybody can post his problems with PhotoLine
wmjordan
Mitglied
Posts: 67
Joined: Wed 14 Apr 2004 16:33

What's the differences between the scaling algorithms?

Post by wmjordan » Tue 12 Apr 2005 16:33

I see that there are several algorithms to scale a layer, like bilinear, bicubic....

What are their purpose to scaling (some good for scaling big pictures and some suitable for small ones)?

Michael Roek-Ramirez
Betatester
Posts: 499
Joined: Tue 19 Nov 2002 16:16
Location: Darmstadt/Puebla/Shanghai

Scale up

Post by Michael Roek-Ramirez » Tue 12 Apr 2005 17:13

Hi WM:

The scaling algos reveal different results when blowing up images.
Scaling down does not make a visible difference IMHO.
The problem blowing up images is, that you generate additional pixels out of existing information of the surrounding pixels. The algo just defines, how those new pixels are calculated.
I cant tell yet, if one of the routines is slower than the other ones.... up to now I just found out the - depending on your image - one or another algo does better!


regards from Mexico

Michael

deli
Mitglied
Posts: 24
Joined: Thu 14 Apr 2005 16:02

Post by deli » Thu 14 Apr 2005 19:24

In case of downscaling it does make much difference, if you have an Image with a lot of noise or grain (scans of high speed negatives or slides).
IMHO the cubic algo does the best downscaling in those cases. It gives a "weaker" (but sharp) image.

Michael

Michael Roek-Ramirez
Betatester
Posts: 499
Joined: Tue 19 Nov 2002 16:16
Location: Darmstadt/Puebla/Shanghai

Okay

Post by Michael Roek-Ramirez » Fri 15 Apr 2005 15:29

Hi Deli:


Thanks for the input!
I did never bother when downscaling, but next time I will give it a try!

A I mostly have problems when upscaling, I observed the behaviour in that case mainly!

Regards

Michael

wmjordan
Mitglied
Posts: 67
Joined: Wed 14 Apr 2004 16:33

Post by wmjordan » Fri 15 Apr 2005 16:43

Oh, thanks for your advice!!

But actually I don't tell too much from those bilinear or bicubic from each other... except that the "Quick.." method which yields the worst result.

Novice23
Mitglied
Posts: 3
Joined: Sun 22 Oct 2006 12:02

Post by Novice23 » Sat 28 Oct 2006 06:08

About a year ago, I conducted some tests on downsizing my photos. The reason I did the tests was that I had just joined Flickr and I wanted to upload some of my pics in a reduced size to conserve the bandwidth I was permitted in a free account.

I noticed after downsizing that many of my pics showed artifacts such stairstepping, or the zig-zag appearance of diagonal lines.

I recall in the tests I conducted then, that Lanczos produced the fewest artifacts. As I recall, Lanczos produced slightly softer results than bicubic, and an additional sharpening was required after downsizing.

I read recently that the newer Lanczos 8 applies a bit of sharpening. I think last year when I did the tests, I used an earlier version of Lanczos in IrfanView. I have not tried Lanczos 8.

The only downsizing I have done recently has been in my RAW converter which produces better results than downsizing after conversion.

I recall reading in tests from others that they found Lanczos to be better for downsizing than upsizing. For upsizing, they preferred bicubic and other methods.

User avatar
Hoogo
Betatester
Posts: 3865
Joined: Sun 03 Jul 2005 13:35
Location: Mülheim/Ruhr

Post by Hoogo » Sat 28 Oct 2006 09:17

Some comparisons for rotation and some algorithms: http://www.path.unimelb.edu.au/~dersch/ ... lator.html