Photoshop CS5 - how to "normalize" colors in an image to reduce the noise?

Question

New to photoshop so bear with me...

I have a scanned image that, ideally, should be just three colors: red, black, and white. The original was a pen an ink drawing and the act of scanning added all sorts of slight color variations.

To make things worse, the image was saved as a .jpg before it was sent to me, so the compression has created some artifacts as well.

I've already tried "reduce noise" filter in a hundred different ways and, while somewhat helpful, does not produce the effect I'm looking for. Is there some way to convert, normalize, or blend all of these different shades of red to a single "red" color? Likewise for the blacks and whites. I don't want 35 different shades of light gray around the edges of the my image.

Thanks in advance for any help you can provide.

Answer 1

There are three "brute force" approaches you can take:

• Add a "Posterize" adjustment layer and set the color count to 3. One you have it to your satisfaction, flatten the image. This is probably the quickest and simplest approach, and will likely work fine for your purposes. This is the only approach that doesn't mandate making a copy of the base layer first.

• Use Filter > Blur > Surface Blur to smooth out the variations without affecting edges.

• Eliminate the noise on a channel-by-channel basis using Surface Blur or Filter > Noise > Median. You can even go into Lab mode and run the noise filter on the a and b channels, then Surface Blur on the L channel, but in a simple case like this I doubt that's needed.

If the jpeg artifacting is really obnoxious around your hard black/white boundaries, as it sometimes is, you can prep the image for the Posterize treatment by using a hard-edged white brush in Overlay mode (or the Dodge tool set to "Highlights" and about 50% exposure) to turn the gray blocks white without affecting the black.

Answer 2

Usually hand-drawn/painted illustrations aren't going to have a single shade of a particular color, even if you used a single color to draw it. That's because there's texture to the paper, to the pen/paintbrush, and there's usually variation in the amount of ink or paint deposited on different parts of the paper. The scanning process itself can also introduce noise, but even with the best scanners, you still won't get solid perfectly uniform colors.

That's why digitizing an illustration is a manual and often very laborious process.

If you have a digitizing tablet, this can greatly speed things up. You can either import the image into Photoshop and digitize it as a raster illustration, or you can import it into Illustrator and make it a vector illustration. It largely depends on your preferences and the style of illustration.

If you're lucky, and you have a very high quality/high resolution scan and an appropriate design, you can use Illustrator's live trace feature to convert your image into vectors for you. This process works better for some images than others. If you just have solid, well-defined regions of color (no gradients) and not too many small details, then this method works very well and requires very little manual tweaking afterwards.

In Photoshop, you can use the magnetic lasso tool to manually select each color region and fill it with a single color (preferably as a new fill layer). Or you could just use the paintbrush tool and eraser and re-create the illustration.

In Photoshop you can also select a color range, but that doesn't work too well in my experience (especially compared to the color range selection in the HSL dialog).

Because it's such a tedious process, I think most illustrators prefer to just sketch the outlines on paper; scan it; and then do the inking and coloring digitally.