User:Jhellingman/Processing Graphics

My current procedure for scanning and processing illustrations is as follows.

Scanning

I scan at 600 dpi, using as many bits as the scanner software allows, which is typically 16 bits per channel. I scan without descreen filters switched on and leave most settings of the scanner at the default.

When scanning, I normally, especially with thin paper but a sheet of black plastic behind the page to avoid the back of the page shining through.

I scan both black and white and grayscale (rastered) images as grayscale.

Even when the scanner does not support 16 bits per channel, I normally use 16 bits per channel while manipulating images in the following steps.

Crop and Rotate

Using Photoshops crop tool, I carefully crop the image, and at the same time take care to correct any rotation introduced by scanning. I typically include any decorative borders present as part of the image.

Remove Paper Texture

The next step is to remove the texture of the paper, using the levels tool in Photoshop. This will make the background appear white, instead of slightly gray in most scans. This requires some careful balancing between removing too little of the paper background or loosing light details of the image.

The levels tool shows a small histogram. Using some sliders, you can select which levels of gray should become white. At this time I also pull the darkest color present in the picture to black.

Remove stains and scratches

Foxing stains

Using the lasso tool, I select the foxing stain. Then with the feather option make the boundary 'soft'. After this I use the levels tool to reduce the stain, while keeping the detail inside it. It may take several iterations of this process, with ever smaller areas to completely remove the stain.

Scratches

Scratches can often be removed using the rubber-stamp cloning tool on an appropriate region. Take care to select a magnification factor and tool size compatible with the features in the image you are trying to remove the scratch from.

Remove legends

I typically remove legends from images if they can be easily reproduced in HTML, and are not integrated in the illustration in some way.

Reduce to Eight Bits per Channel

At this point, I reduce the image to 8 bits per channel, and save it as a TIFF file. For long term storage, eight bits per channel is normally sufficient. From now on, this file will be the master version for all further editions.

Create Low Resolution Versions

For my HTML versions, I typically create an image that fits in a 720 by 720 pixel box. To do this, I perform the following steps.

1. reduce to 144 dpi.
2. reduce such that longest side of image is 720 pixels.
3. make darkest part of image 10 steps less than black in levels tool
4. use unsharp mask to increase sharpness (33%)
5. save as jpg (photographic images)
6. save as jpg, png, or gif (whatever is smallest) (drawings)
7. save as png or gif, using just 8 levels of gray (line-drawings)

Whenever an image requires more detail to convey all information in it (such as maps, etc.) I will also prepare a high-resolution version of the image, and make the image link to that high resolution version.