Optimum scanning resolutions for images in Sigil?

[Serpentine]

Some random stuff off the top of my head:

PNG - black/white, make sure what you're saving either has a greyscale/b&w/palette (8 bpp) option and use it.

Don't use GIF, use 8bpp/paletted PNG. While you can save a few bytes sometimes, it's not worth it.

Screenshots and such should be PNG, nearly all of the time you should be using a palette/8bpp. If you use services like smush.it, they will not try to convert for you, so make sure you save correctly the first time.

Diagrams - Where you have black/colour on white, it's often nice to make the white areas transparent, allowing readers with a background/page texture to not be harshly excluded. This however is seldom an option in vanilla editors, look for a 'white to alpha' or 'colour to alpha' plugin. I find the ones for Paint.Net quite good, Photoshop can be done easily enough with masks (but it's :effort.

JPEG - If you're using photos, I strongly recommend using something like XnView, you can often get quite a dramatic decrease in filesize while preserving details by playing with the subsampling factor in conjunction with the quality.

Don't posterize full-colour images to save space, e-ink has 16 levels of grey, but that doesn't mean your book wont be read elsewhere. The built in dithering is better anyway.

Remove your metadata, really - all of it. Colour profiles are actually used in some firmware, since the render packages often have defaults enabled, this is slow and generally useless. I've seen 20kb jpegs with 150kb+ of metadata/profiles.

If your book has many images in a single section - be very careful with using high res, iThings or not. A lot of newer readers (including eink) will use full RGB(A). Your 1k x 1k px image might only be a 30kb file, however it's 4.2mb in the readers memory. If you double the res, you quadruple the pixel count (2k x 2k = 16mb+ ). Not a problem for covers and such. But for sections full of pictures, you end up hitting performance hard on lower spec readers, and even on more beefy ones, slowing things down.

SVG - Tools like SVG Cleaner often save you some entities and such, which is a very good idea.

[dwig]

Originally Posted by Jim the Obscure

...

My question is (& I've searched a lot) what is the optimum PPI resolution for scans of b&w photographs to be imported into Sigil for display on both e-readers and in something like Calibre on a computer? ...

In actuality, there are two answers.

First, to ensure that you get the best results you should ALWAYS scan photographic prints at something in the range of 300-600ppi. Film should be scanned at much higher resolutions; 35mm at around 3200-6400ppi, 120 at around 1800-3200ppi. This PPI is the scanning resolution and is measured relative to the physical original placed in the scanner. These files should be saved in an 8bpp color format that does not use a lossy compression method (read: avoid JPEG like the plague and don't use PNG unless you really really know what you are doing and know your scanning application EXTREMELY well). You can scan directly into a good image editing app to avoid saving a disk file at this point, though saving a high quality "master" file at this point is a good idea in case Step 2 (below) has to be repeated.

Second, after scanning to acquire what is reasonably all of the data in the original, you should downsample to the appropriate pixel size. Don't be concerned with PPI at this point. In fact, it is best to avoid saving any PPI value in the final file (e.g. by using Photoshop's "Save for Web & Devices..." option). Use a high quality tool to do the resizing and ony after resizing should you convert to the desired format; JPEG, GIF, PNG.

PPI in a digital image doesn't really exist other than as a note in the file's header that tells those applications that have their own concept of a virtual inch how the creator of the image wants the pixels to be scaled. A 300x300 pixel image tagged at 100ppi has exactly the same resolution (detail, sharpness, ...) as a 300x300 pixel image that is tagged at 300ppi. The only difference is that some applications will deflaut the image sizing on import to the spec'd PPI resulting in the first image being scaled to 3"x3" and the second to 1"x1".