How to make gritty two color photo?

Question

Reducing colors with levels will give you very high contrast stencil-like results. Half-tone effect will be too "perfect", stamp effect gives too much contrast too. Any ideas how to make similar images? Diffusion Dither is different thing.

Answer 1

There's a free plugin called G'MIC - which works on Photoshop, GIMP and Krita, and possibly some other image editors too.

The G'MIC Charcoal filter comes close. It's located under the Black & White filters in the plugin. The controls can be tweaked.

An example

Answer 2

I propose using the Dissolve blend mode to create that gritty look.

As example I'll use this image which I've corrected to be pretty light without too large areas of black.

^{Photo by Paul Brennan, CC0 Public Domain}

^{Closeup at 400%.}

• Copy image to clipboard.
• Make a white layer.
• Make a black layer.
• Make a mask for the black layer, paste in the image and invert it.
• Set the Blend Mode of the black layer to Dissolve.

The result looks great, but it's harsh 1-bit and the grains are smaller than what you are looking for:

^{Click image to see full size.}

^{Closeup at 400%.}

Merge the black and white layers (or turn them into an Smart Object) and apply a Gaussian Blur filter with a Radius of 0.5 px for a more appealing result:

^{Click image to see full size.}

^{Closeup at 400%.}

I really like this look. Grainy but smoothened a bit. Perhaps it could be of use to you.

To make the grains coarser like in your examples, apply a Dust & Scratches filter with Radius set to 1 px before you apply the Gaussian Blur filter:

^{Click image to see full size.}

^{Closeup at 400%.}

All dithering methods like this are very dependent on the resolution of your image. If you feel you lose too much information, consider applying the effect at double size and then scaling down to half size afterwards. This will give you smaller and sharper grains:

^{Click image to see full size.}

^{Closeup at 400%.}

Answer 3

Make a DIY-halftoning based on noise. Start from an RGB photo which contains black and white and there's also plenty of intermediate brightness levels. The photo can be as well colored as desaturated. It's pixel dimensions should be selected so that the actual pixels are about the same as the screen pixels. The next example is 1200 px wide:

Add a new layer. Fill it with 50% bright grey. Apply to it Filter > Noise > Add Noise > 100% Uniform Monochrome

Blur the noise. Apply Filters > Blur > Blur. Gaussian blur can be used if the radius is less than one pixel. Bigger radius makes the grains too big in next steps.

Adjust in the Layers panel the opacity of the blurred noise to 50%. Insert a tresholding adjustment layer and adjust the treshold. The half-toned greyscale is not especially linear, so you must adjust the treshold for good tones in important areas like on the face:

Inserting a curves layer just above the photo can help in some cases. In the next example the midtone contrast boost with a curves layer makes the tones in the face substantially better:

Tresholding should generate only black and white pixels, but you may see grey in the screenshots. That happens because the image is resampled at least when one watches it in the web browser. In Photoshop it looks right when the zooming is set to "Actual pixels".

You may want less random grains but you do not want the same degree of regularity that one gets by converting the image to grayscale, then to bitmap mode and by applying "diffusion dithering". As a compromise you can make the noise pattern from flat 50% grey by applying in separate image to it diffusion dithering, Convert the result back to RGB, add some noise and blur. Use it as your blurred noise layer:

The mechanical looking diffusion dither pattern is much reduced by applying 50% noise to it before blurring. But the grain size variation is smaller than in pure noise versions => the image looks cleaner.

The blurred noise layer can be also scaled to bigger size for bigger grains. In the on diffusion dither based version it doesn't render the resolution down as steeply as in the pure noise version.