It is not very rare that you want to use/copy an image from a research paper. The easiest way is to just take a screenshot and crop accordingly. But that is not the wisest way. There are two problems with this method:
1) You are creating a raster image (instead of a vector image)
Raster (bitmap) images are made up of individual colored squares, known as pixels. Vector images, on the other hand, are made up of lines connecting two dots by a computer algorithm, instead of just a line of pixels. The difference is emphasized when you zoom them.
Screenshots are by default raster images. Hence, improper crop dimensions or improper stretch while using them can cause weird distortions.
2) You cannot crop overlays
It is common to have letters overlayed on the objects, to refer to them in the description. But you may not want them. There is no easy way to crop them out.
The wise way!
Below the PDF download option in arXiv.org, there is an option to download the source of the paper. This is usually delivered as a gzipped (.gz) file. This will contain the tex file used along with all the images used. These images are in pdf file, a vector image file format. They can be easily converted to svg either using inkscape:
inkscape -l out.svg in.pdf
or using pdf2svg:
pdf2svg <input.pdf> <output.svg> [<pdf page no. or "all" >]
Once you have the svg file of the image, you can just open it in Inkscape (or any other vector image editor). Now just delete unnecessary objects and export it to png (or whatever you prefer).
Done! Enjoy your resizable, selective copy!
Don’t forget to cite the images properly.
Download the source from arXiv. Convert the images to svg. Edit in Inkscape. Export to png.