Ah, sorry the translation is a bit confusing—but to be fair it is a slightly confusing topic as well, even for those that know a fair bit about graphic design. Well to separate out a few different threads:
- The “DPI” of an image is simple a ratio between the configured display size of the image (cm, mm, in, pt…) and however many pixels that image has stored within it. If an image has 300 pixels per inch and it is 5" wide, then that image is 1,500 pixels across.
Importantly, there really is no such thing as DPI, in the sense of there being a setting that needs to be protected or some-such. It is the result of a mathematical formula. So long as the quantity of pixels in the image does not change (which Scrivener will not do) and the display dimensions on the page do not change (which Scrivener will not do) then the “DPI” is the same. - Consequently has nothing to do with file format or compression—so long as the file format supports the notion of storing display size of course. I believe for instance GIF does not, being designed long ago purely for digital output.
I don’t
think for PDF Scrivener needs to do any processing or recompression. If it does, JPG files will convert to PNG files internally during compile, which means they get converted back to JPG. In accordance with the above, that doesn’t change anything with regards to the resolution of the image, but for highly compressed images it may incur a drop in quality. For one iteration like that it’s not going be something you’ll even notice in most cases. You would have to analysed the image at the pixel level in a tool like Photoshop to even see where the recompression impacted the image. Recompression is generally something you want to avoid, but more as a side-effect of iteration—i.e. you wouldn’t want to take the image from the PDF and put it back into Scrivener and compile again, and then repeat a dozen times. Continually recompressing from the same source file however is fine.
The DPI setting in Preferences is purely for vector graphics in Scrivener that must be converted to raster images because the output format does not support vector. It determines how many pixels per inch will be used.
I would also add the general disclaimer that I don’t claim the macOS print driver’s PDF generator is the best place to be typesetting a final book with production ready graphics. I’m not sure how that is going to handle colour correction, profiles, CMYK, etc. This is really more meant to be a facility for proofing copies than replacing a design environment. We cannot vouch for these processes being used to create professional results.