Image Resolution Tips for Journal Submission

[Urmi Roy]

Hi,

My paper has been accepted by a journal. However, I have a problem with the images...

The reviewer says:
"it seems that the authors are importing or inserting .png or .jpeg format pictures (Figure 2, 4 , 7, 8, 9 , 10, 11, 12). I would recommend saving them in a vector graphics format first (eg. .eps format) at atleast 300 DPI resolution, and then importing them to gain better clarity."

I managed to make the images into jpeg images at 300 dpi. However, after converting these into .esp files, my computer (Windows operating system) can't open them. Also, making them into .svg files distorts the images.

Is that enough or do I try something else?

 

[bigfooted]

How did you make the images? The reviewer thinks the resolution of the images is not high enough for publication. If it is being printed, the images will look blurred.
When you create an image consisting of geometric shapes like lines, squares, circles, etc in whatever program, you usually have the option of actually saving the lines, squares, circles, etc. instead of immediately converting it into pixels. Then, when you scale the image, it will never get blurry because the program will just change the radius of the circle or the length of the line. The most common vector format is postscript (ps) or enhanced postscript (eps). You can use an image program like gimp to view ps files and convert it to other formats. You can always convert postscript to jpg, but once it's jpg, you cannot convert the jpg back to postscript - the vector information is gone. Hope this helps.

 

[Urmi Roy]

Hi bigfooted, thanks for your reply. Most of them were initially created on MATLAB, and some on another computational software called COMSOL. The problem is that it'd take years (and loads of energy) to re-make the diagrams and store them as ps (since you said jpeg images cannot be converted to ps).
However, as I said, I did make them into 300 dpi using my mac. Will that not be enough? So if I can't make my existing images into .esp, is there any other format I should try? I tried .svg, but my image gets really distorted...

 

[bigfooted]

Don't you have the computational results anymore from your MATLAB and comsol simulations? When you still have the raw data and when you create figures with matlab, you can save the MATLAB figures as postscript files with the 'print' command. When you still have the comsol results, you can just take an image snapshot and change the dpi and x and y pixel size of the image.
And how did you create the 300 dpi images you are talking about? You cannot do it with only the original jpeg image, it will just scale the bad quality image to a larger bad quality image.

 

[Urmi Roy]

Hi bigfooted, unfortunately I can't find all the data for all the images...there are so many images and they were altered so many times...about the new 300 dpi images...well, my dad has a macbook, and I opened my ppt on keynote on his mac. I then copied the images from keynote onto a program called 'preview' and then saved the images as jpeg files with higher resolution.
I'm attaching a document with one of my figures, in 2 different versions...pls let me know if you can see any difference (the new 300 dpi files are much bigger in size).

 

[physorguser]

Hi,

My paper has been accepted by a journal. However, I have a problem with the images...

The reviewer says:
"it seems that the authors are importing or inserting .png or .jpeg format pictures (Figure 2, 4 , 7, 8, 9 , 10, 11, 12). I would recommend saving them in a vector graphics format first (eg. .eps format) at atleast 300 DPI resolution, and then importing them to gain better clarity."

I managed to make the images into jpeg images at 300 dpi. However, after converting these into .esp files, my computer (Windows operating system) can't open them. Also, making them into .svg files distorts the images.

Is that enough or do I try something else?

I'd forget about eps or other now arcane formats. If jpegs themselves are not good enough and I fail to see why that would be then the best universal format to use and why it was invented is to use a pdf. A pdf is not a file format as such but it is a container, it can contain vector or graphics, if you were to save a jpeg as a pdf from say Photoshop it would still be a jpeg but it would be contained in a pdf file. At least you know that anyone should be able to work with it in that format.

As far as resolution goes you do not want to be up sampling because you will just reduce the quality. The resolution by itself means nothing without knowing the size. Best just to set the size that you want the image and let the resolution fall where it may. If there's too much resolution it will get thrown out and if there's not enough well there's not enough, you can either live with it depending on how important the size is, for example 120 dpi should still be enough for most purposes, but if it's something like 50 then you'll just have to make the image smaller.

It is easier to do this if you can supply the exact sizes you need and the pixel size of the image and the destination size and output.

 

[Borek]

Converting and rescaling images that are already bad won't make them any better, no matter how long you will try. It is a waste of time.

IMHO doing them from scratch and exporting in high resolution (better yet: in a vector format) is the only option.

 

[Urmi Roy]

Hmm okay, I get the idea. Since my article has been chosen for an elsevier journal, I've asked for professional artwork assistance that is provided by elsevier. Let's see what happens!

 

[AlephZero]

unfortunately I can't find all the data for all the images...there are so many images and they were altered so many times.

So far as the content of the paper is concerned, that is probably a bigger problem than the quality of the images, if you don't have an "audit trail" from the original computer models to what you are publishing!

 

[Urmi Roy]

@AlephZero - Yeah, it's turned out to be a real handicap. I never thought I'd have to recreate the images, so I didn't particularly take care, and this is my first paper. This is a good lesson for next time though!

 

[Urmi Roy]

I got to redo the images, since there is no other way...I re-did the plots on MATLAB. I now have an option to save them in .eps (which my computer can't open) or in.emf (metafile, which my computer can open) version. Which would you suggest?

 

[D H]

Get the right conversion tools and you should have no problem looking at an encapsulated postscript (eps) file. There are plenty of eps to pdf converters out there.

Rather than redoing the plots in Matlab and exporting the plots as an encapsulated postscript file, you might want to consider exporting the data that form the basis of those plots from Matlab as a text file and using some other tool to generate the plots from those saved data.

 

[Urmi Roy]

DH why do suggest using some other plotting tool over MATLAB? I did some of the plots on MATLAB, and saved them as .eps, but there was no means to specify the resolution to 300 dpi. Will that be a problem you think?

 

[Urmi Roy]

Oh and another problem is that since COMSOL is used to make simulations, the images are pretty complex. When trying to export the images, it allows us to save only in .jepg or .png formats...What should I do about this?

 

[Borek]

saved them as .eps, but there was no means to specify the resolution to 300 dpi

eps is a vector format, it doesn't have a dpi resolution.

Well, it can be used to save rasterized images, but that's not the situation we are talking about.

 

[D H]

Some people think Matlab can produce publication quality graphics. My opinion differs. Those Matlab plots typically are not what I would call publication quality. Matlab produces splashy graphics. Publication quality graphics are rather austere. The background should be white, there shouldn't be any grid lines, the legend (if any) should be suitably placed and not be a distraction, and the fonts are consistent with those used in the main body of the document.

Even if you overcome the first three issues (background, grid lines, legend), the last issue (font) is very hard to address with Matlab-generated plots. That encapsulated postscript file produced by Matlab is almost inevitably going to call for different fonts / different font sizes than those used in the main body of the document.

Elsevier is going to help you, possibly with a surcharge, on your graphics. If you give them the Matlab-generated plot only, about all they can do is edit the .eps file to change the font, and even that can be problematic. If you give them the plot and the underlying data, they can feed that data to a tool better suited to creating publication quality graphics. Now your Matlab plot is just a guideline for how the plot should finally appear.

Alternatively, you can sidestep this Elsevier dance and use a tool better suited to publication quality graphics yourself. R, gnuplot, TikZ, etc.: There are a boatload of tools that do a far better job of producing publication quality graphics than Matlab.

 

[jim mcnamara]

R and gnuplot are no-cost downloads, if that is a factor. TikZ is also available free from sourceforge.net -- it is messier to user, IMO -- unless you are already good with Tex. I'm not.