A graphic is worth 1000X

Derrick Rohl, who is currently doing astronomy research in Chile, sent me an email about LaTeX, the program he uses to write scientific research papers. Toward the end, he mentioned a feature he had recently discovered:

I didn’t realize [LaTeX] could do plots. It may be a while before I need to use them at all, but I noticed the labels and numbers on axes could be highlighted, rather than being part of a jpeg.

An astute reader of my blog might be taken aback by that last bit – namely, that using a JPEG image would be a good way to include a graph in a document. Why? Because, just last week, I glazed over some of the pitfalls of JPEG compression. Instead, I said, he should use a vector-based file format, such as EPS, to save images of this sort.

What’s the EPS advantage?

The issue is how the image is stored in the file. Vector-based formats, like EPS, SVG, PDF, etc. basically store a set of mathematical equations describing the image, whereas rasterized formats, like JPEG, PNG, GIF, TIFF, etc. store a pixel-by-pixel representation. Because vector-based formats store mathematical representations of the shapes in the image, they can be infinitely resized without losing quality. An EPS version of a line graph could be a fraction of the size of a TIFF file representing the same graph, but the EPS will not become pixellated, even when zoomed in 1000X.

What about JPEG?

JPEG, specifically, is a lossy compression format. If you open an image in Photoshop and save it as a JPEG at maximum quality without changing it at all, you will have lost some detail in the image itself. That loss of detail is barely noticeable in a photograph, but computer-generated graphics make JPEG artifacts much more obvious because they’re simpler images to begin with. If you absolutely cannot use a vector-based format (for instance, some companies don’t publish vectorized copies of their logo in order to protect their brand), you should export to PNG, not JPEG.