Re: [PanoToolsNG] I need a test image to show the difference between Lanczos and BiCubic iterpolation...
- Hi Fulvio,
There are a bunch of different bicubic interpolation methods,
kernals, basis functions, polynomials, or whatever you call them.
The one I use -- and I don't know how to identify it -- has
one variable that I can change to make the basis functions produce
a sharper or softer interpolation.
I've read that either Lanczos or bicubic interpolation is better
for upscaling or downscaling but I forget which one is best applied
But at the resolution we work at on the web -- typically about
1k x 1k viewable and 8 bit per channel color -- I don't think it
really matters which method is used. Except for the ringing which
can be seen if sharpening is done too agressively.
For print, I'm sure that the differences would become apparent immediately.
There's other interpolation methods which I know nothing about except
that they do edge detection and interpolate the image in a way that
takes edge detection into account. The examples I've seen
look very good. I think I've seen the acronym NEDI associated
with those methods.
Fulvio Senore wrote:
> You can see some example in this page:
> there are bilinear and some Lanczos ones, no bicubic, sorry.
> Bicubic is a rather generic term, but some time ago I looked at the
> Lanczos and bicubic implementations in PTGui and I believe that bicubic
> normal is better.
> Lanczos has a slightly sharpening effect that can cause ringing (echoes
> of sharp edges, like in overcompressed jpegs).
> For this reason IMHO Lanczos is better for viewers where a sharp image
> is more pleasant, but bicubic is better for file creation, where the
> lanczos oversharpening might not be desired.
> About the interpolators test in the Dersch page, I think that it is not
> representative of real usage. Nobody applies so many interpolations to
> the same image.
> That test shows that a large windowed sinc kernel works very well, but
> this is not always true. Just try to interpolate a screenshot with sharp
> edges and you will see large ringing effects.
> Fulvio Senore
> Ken Warner ha scritto:
>>I've written my own Lanczos and bicubic interpolators. When I apply
>>them to the typical equirectangular pano, I really don't see much difference.
>>I'm trying to talk myself into believing that the Lanczos interpolator is
>>better than the bicubic interpolator but I can't. I just can't see the
>>difference. In fact I can barely see the difference between bilinear
>>and bicubic interpolation. Bilinear is only slightly softer.
>>So what I need is a test image that I can use to show the difference
>>to myself. Dr. Dersch used some images with concentric circles --
>>Where does one get a test image like that?
>>Or does anyone know another image that will show defects and clarity