From: rudolf D?lling on
I need an improved representation of measured data in a contour plot.

The data are given on a regular 2D-grid. Due to the measurement procedure the grid pitch is dense in x and not dense in y. From the underlying physics it is clear that the distribution has a single peak.

You can see more clearly what troubles me with the "contourc" algorithm when looking to the following data field:

0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
0, 1, 2, 4, 5, 4, 2, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0
0, 0, 0, 0, 0, 1, 2, 4, 5, 4, 2, 1, 0, 0, 0, 0, 0
0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 2, 4, 5, 4, 2, 1, 0
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0

The contour lines from "contourc" will depict three hills here. A representation better fitting my needs would be a single long straight hill with a flat ridge from above left to below right. This would need an algorithm "looking for equipotential values at neighbors which are farther away in x but close in y". This non-isotropic algorithm would be of course more complex and probably more sensitive to noise than "contourc". The algorithm could be something similar to a "contourc" applied to "convhull" in 3-D space, but the measured data are usually not convex and hence to much information would be lost.

Is there a way to produce such contour lines without "inventing" additional data lines with (in a special way) interpolated data?

I would be glad to hear comments.