how do I remove lines that show up in contour plots?

When I save a contour plot created using contourf as a .eps or .pdf file, there are vertical and diagonal lines through the cells. These do not appear in .fig files or other versions of the output. These are ugly and distracting. How do I get rid of them? I am running R2014a. This did not happen in earlier versions of Matlab.

[Jan]

Can you post a screenshot of the EPS file?

[Walter Roberson]

Alan, could you confirm that you are running R2014a, and not any version later, R2014b, R2015a, or R2015b ? This is a known difficulty for R2014b and later, but not for R2014a.

[Alan Chave]

It is R2014b. The problem only occurs for .eps and .pdf, but this site will not allow me to attach either file format

[Walter Roberson]

You can zip any kind of file and attach the .zip provided it is not too big.

Answer 1

[Kelly Kearney]

I'm guessing these are antialiasing artifacts, as discussed in this thread. The problem has been around for a long time, but some recent changes to the way Matlab triangulates all patch objects in the new graphics system means that the problem rears its head far more often in R2014b+.

Answer 2

[Walter Roberson]

http://uk.mathworks.com/matlabcentral/answers/166366-white-line-etching-on-exported-eps-figures-in-r2014b

http://uk.mathworks.com/matlabcentral/answers/169015-r2014b-tesselates-contour-plots-when-saving-to-eps

[Alan Chave]

This raises the same issue that I did, but does not offer solutions. None of the solutions offered in my first post work at all, and I spent much of the day trying them (except for editing the .eps file, as I do not have enough understanding of postscript to do that). My present position is that Matlab cannot handle my graphics needs, and I will have to look elsewhere. The view point that it is not a matlab issue leaves me cold, as something changed from R2014a and later releases that breaks output in .eps and .pdf format.

[Walter Roberson]

Alan, there are bugs in many of the common PDF readers with polygons that are too simple ("Too soft"), and there are bugs in many of the common PDF readers with polygons that are too complex ("Too hard!"). You appear to be presuming that there is a Baby Bear Solution that will be "Just right!" for all of them. But do you know that to be true? The nature of bugs is that the intermediate settings could be wrong for all of the most common PDF readers.

If Mathworks needs to choose which PDF readers to be compatible with, if it is not possible to create output that works with them all, then which ones would you say are mandatory that MATLAB works around the bugs of, and which ones would you say it is okay for the MATLAB output to not work with?