5

I am currently writing my masters thesis and I created the following graphic:

enter image description here

While it looks ok in color, the areas are not so easy to distinguish if this was printed on a non-color printer:

enter image description here

Do you know any 6 colors which are suitable for this type of image and which can be distinguished if printed with a non-color printer?

(Other ideas how to make sure the reader knows where the numbers come from are also welcome, but I guess colors are the easiest way to do this.)

6

Well. In reality your colors could work well. The problem you have is who is converting the colors.

Your grayscale image looked totally strange to me because the yellow should be printed very light.

Here is your image converted to grayscale with a simple editing program. (The black background is just because it is a transparent png)

enter image description here

There is a chance that whatever "software step" it is converting your image is making a strange decision.

It could be a printer driver, or something.

There is a chance that that driver is only taking in acount saturation. All colors look saturated. Yellow, magenta, cyan, green, except the pink that could be an unsaturated red.

You could try to use an unsaturated yellow (pastel yellow), the unsaturated red( pink), a saturated cyan, etc.

But probably what you really need is to make diferent versions or the file if you are using a non profesional program for making your publication, and you need to stick with it.

If you are using any profesional program the color mode conversion will be more predictable. You only need to export to a PDF with a grayscale color output.

You could use: Scribus https://www.scribus.net/, Page Plus Starter edition http://www.serif.com/desktop-publishing-software/ which are free.

But even Word makes a better color conversion, this is a pdf generated from PDF creator, which is a driver to make a pdf printing from any aplication. It is pixelated because I just made a quick conversion, but the grays are better.

enter image description here

In any case, you could print this kind of images: https://www.google.com/search?q=rgb+color+swatches and choose some that fits your overall design in color and in grayscale. But again, I would use another program to prepare my publication.

4

I suggest you try Colorbrewer. It will let you pick a scheme and add choices as to colour blind friendly, print friendly. It might not be the perfect tool, as it is for maps. Therefore, its criteria is that all colours must work in contrast with each other. You need something simpler, so it might well work perfectly, even if Colorbrewer says "no such scheme". Ignore that, and try printing something.


EDIT: This is a "gradient of one colour". Of course this would work; and it would print a reasonable grayscale.


Actually, i think the best answer is here Selecting colors suitable for color, greyscale, and black/white printing

enter image description here

4

Suggestion: you can do away with the colors altogether. Instead, just emphasize the blocks in the original matrix, and retain the emphasis in the max-pooled matrix. The max-pooling process is simple enough that your readers should easily be able to see what's happening, especially given the context of the rest of the document, and you don't have to rely on the mercy of your printer drivers or your readers not having color vision deficiencies.

Here's an example of what I mean. You can change the very thick thickness to something else if you feel that it's too thick.

Sample image.

\documentclass[tikz]{standalone}

\begin{document}

\begin{tikzpicture}
  \def\margin{2cm}
  \tikzset{block grid/.style={very thick}}
  \begin{scope}[xshift=-6cm-\margin,yshift=-2cm]
    \draw[help lines] (0, 0) grid (6, 4);
    \draw[block grid] (0, 0) grid [step=2cm] (6, 4);
    \foreach \row [count=\j] in {%
      {9,2,9,6,4,3},%
      {5,0,9,3,7,5},%
      {0,7,0,0,9,0},%
      {7,9,3,5,9,4}%
    } {
      \foreach \cell [count=\i] in \row {
        \node at (\i - 0.5, 4 - \j + 0.5) {\cell};
      }
    }
  \end{scope}
  \draw [|->] (-0.5, 0) to node[above] {max pooling} (0.5, 0);
  \begin{scope}[xshift=\margin,yshift=-1cm]
    %% Optional: uncomment the following line to draw mini-grids inside
    %% the smaller grid cells, to really reinforce that each cell here
    %% comes from one block in the grid at left.
    % \draw[opacity=0.2] (0, 0) grid[step=0.5cm] (3, 2);
    \draw[block grid] (0, 0) grid (3, 2);
    \foreach \row [count=\j] in {{9,9,7},{9,5,9}} {
      \foreach \cell [count=\i] in \row {
        \node at (\i - 0.5, 2 - \j + 0.5) {\cell};
      }
    }
  \end{scope}
\end{tikzpicture}

\end{document}
  • Note that using heavy rules—or vertical rules at all—in tables is very rarely a good idea. (Cf. texdoc booktabs for more info.) However, I don't really think that it's a problem here, because it's not a formal (data) table, and the rules are semantically meaningful. – wchargin Nov 20 '16 at 15:26
0

The general answer to the question as you asked it is to start by choosing your colors in a color space that has a lightness or brightness component (or similar), and ensure that there's a substantial difference in the brightness/lightness component.

At least assuming your printer driver does a decent job of converting color to grey scale (something that's been well known for many years) this should do a good job of ensuring that the patches are easily distinguished after grey scale conversion.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.