Scanning b/w images and diagrams

[SkookumPete]

This seems like a question that many must have asked, but I can't find any information on it.

For my current project, I am scanning (as images) several diagrams and maps that appear in black and white in the original book. I want them to have a solid "white" background and of course I want the text, which is quite small in some cases, to be sharp and readable on the Kindle.

Now one would think the procedure would be simply to scan the images in black and white. However, I've tried this and the results are not satisfactory. For one thing, the text is not anti-aliased and becomes unreadable when scaled down. Scanning in grey-scale has presented other problems.

So, as complicated as it may seem, here's the method I've been using, which does produce good results.

1. Scan the image in full color at 600 dpi and export to Paint Shop Pro.
2. Straighten and crop the image.
3. Using "smart photo fix," select white and black areas on the image to correct the color balance. (This step may not be necessary, but my scanner tends to interpret slightly yellowed paper as orange.)
4. Using the "magic wand" to select by brightness, select all dark pixels by clicking on a black area (first adjusting the tolerance so that darker shades of grey get selected too). Then adjust the brightness of the selection to darken the print and fine lines. I've found that adjusting by -80 or so works well, retaining the grey anti-aliasing.
5. Invert the selection so that all light-colored pixels are selected. Adjust the brightness all the way up (255) so that the background turns pure white.
6. Resize the image to the Kindle maximum of 600 pixels wide or 800 tall.

I'd be curious to know if there's a better technique that yields nice dark images (similar to those of the Kindle screen savers) without sacrificing the readability of small print.

[Jellby]

1. I scan in grayscale, or in color and immediately convert to grayscale.
2. Straighten and crop.
3. Add black layer on top, in "blacken" mode. (This makes all non-white pixels appear black, white pixels remain white).
4. Adjust the levels in the bottom layer, until the background is white (visually), and the black is black (according to the histograms).
5. With the eraser, or a white brush, I delete all specks that remain in the background.
6. Remove the top layer, resize, sharpen, etc., depending on the image.

[DSpider]

You should look up SVG (Scalable Vector Graphics). It will do exactly what you want.

Inkscape, apart from being the ultimate open-source, multiplatform, SVG editor, it can also trace images:

http://www.youtube.com/watch?v=rzmdqwFpxp8
http://www.youtube.com/watch?v=wFQ8Y386DWo
http://www.youtube.com/watch?v=r8cpJ...eature=related

You can even add numbers, text, whatever with Inkscape. All in selectable, searchable form.

For a much more accurate trace you could try Vector Magic (and edit the result using Inkscape).

[dwig]

Originally Posted by DSpider

You should look up SVG (Scalable Vector Graphics). ...

Since the OP said this is for a Kindle (his step #6) he should ignore SVG.

OP: You step #1 needs to be rethought. The scanning resolution needs to be chosen based on the physical size of the original and should likely change from original to original if their sizes differ. You need to "over-scan" so that you have more pixels than you need in your final image.

I would suggest scanning so that you initially have an image 4x you final desired size. If 600x800px is your target then scan so that you initially have a 2400x3200 image. If the original is 3x4" you would need a scanning resolution of 800dpi, if its 6x8" you would need only 400dpi. When you downsample latter in the process you need to do so choosing appropriate methods so that you achieve the anti-aliasing that you need.

You may find that replacing the scanned text with new, possibly larger or bolder, text while editing the image will work better than trying to reproduce the original in the scan.

[DSpider]

If the Kindle doesn't do SVG, lmao @ teh Kindle then. Alternatively you could save the (cleaned up) vector graphics as rasterized images, which would give the sharpest diagrams possible.