[
https://issues.apache.org/jira/browse/MATH-1166?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Gilles updated MATH-1166:
-------------------------
Attachment: MATH-1166.patch
I finally tracked down the bug, thanks to web page that showed a code excerpt
from the well-known "reference-that-cannot-be-named".
There were actually two bugs:
* one was indeed a transposition of the coefficient matrix,
* the other was that the implementation on Wikipedia was assuming a _one_-pixel
"delta" between samples.
In the attached patch, I've created new classes:
* {{BicubicInterpolatingFunction}}
* {{BicubicInterpolator}}
The reason is that the fix entailed the failure of the tests in
{{BicubicInterpolatingFunctionTest}} related to the methods providing the
partial derivatives functionality. A solution could have been to remove those
but it cannot be done because of backward compatibility.
The {{TricubicSplineInterpolator}} must be deprecated since it is based on the
buggy code, and the implementation requires the partial derivatives...
A problem remained in that the {{BicubicInterpolator}} needs to estimate the
derivatives in order to use the {{BicubicInterpolatingFunction}}. This entails
that, on the border of the sampling range, the interpolation is wrong. As a
workaround, the method "isValidPoint" is overridden, and for the "valid"
points, the interpolation is within a reasonable tolerance.
This is a weak point of the {{BicubicInterpolator}}, but in some cases (as in
the picture produced by Ajo's code), there could be less artefacts.
Is it OK to commit this patch?
Or do we proceed otherwise?
> BicubicInterpolation may have a bug
> -----------------------------------
>
> Key: MATH-1166
> URL: https://issues.apache.org/jira/browse/MATH-1166
> Project: Commons Math
> Issue Type: Bug
> Reporter: Ajo Fod
> Attachments: Draw.java, MATH-1166.patch, myimage.jpg
>
>
> I plotted a random image with bicubic interpolation hoping to see something
> like this:
> http://en.wikipedia.org/wiki/Bicubic_interpolation
> ... but instead I got something weird.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
