Use GIMP. First, set the image mode to indexed then create a new file with the color space set to grayscale and copy your file to the newly created image.

If doing things incrementally/manually, in general, proceed in this order

a) Scale/resize (compensating for aspect ratio as needed)
b) desaturate and/or convert colorspace and/or dither

Low color and already dithered (low bitdepth color and actual black and white) images scale resize quite poorly.

Not relevant when starting with high resolution and color images, but when resizing/scaling already low resolution and/or color images, it may be advantageous to do so while trying to scale to the nearest possible integer resolution (using integer dividers or multipliers from your source resolution) below your screen target resolution and possibly choosing nearest neighbor interpolation/decimation rather than other scaling algorithms. YMMV

I usually still can meet my all converting retouching work with Paint Shop Pro. 3.11(2) or 4.15SE. Especially palletted stuff.

You'll probably want CompuShow and Deluxe Paint.

I tested some of my files yesterday....can't make quality stuff in the recommended standards...
I will upload the originals so that you can have a try....

Let's talk formats.

PCX can be 16color pal, 256c pal, 24M pal?
GIF can be 2 colors, 4 colors, 8 colors, 16 colors, 32 colors 64colors, 128colors 256colors palette.

GIF is not ideal when convertig for Atari/Amiga, but for PC it's excellent.
What is the real difference betwenn the both gif-formats and the compressions is..I don't know.
Not all programs can save all pal-versions. (2/4/8/16/32/64/128/256)

2 Colors (monochrome) would be ideal for:
- 720x348 Hercules (EuroPC/Tandy1000, Amiga Bridgeboard)
-640x400 Olivetti (AtariST PC
- 640x200 CGA Hires

4 colors
- EGA Hres ?
- VGA Hires ?
- Tandy Hires?

8 colors
- Amiga OCS/ECS

16 colors
- Plantronics (EuroPC) (16pal)
- EGA Lowres/Hires (EGA Pal) 16/64 pal
- VGA Lowres/Hires (VGA-Pal) 256k pal
- Amiga OCS/ECS inkl. Med/Hires 4096pal
- AtariST Lowres 512pal

32 colors
- Amiga OCS/ECS inkl. Med/Hires 4096pal

64 colors
- Amiga OCS/ECS
- EGA Hires 640x350 64pal

128 colors
- Amiga ?

256 colors
- VGA Lowres/Medres/Hires 256k pal.

I rescaled some monochrome art and the output is much worse quality than the original stuff.

I thought about not resizing while the software changes the 2colors to a much higer pal and than does it's thing.
Converting the result back to 2colors looks BAD.

I was thinking about making the image 2x of the target format and takeout every second line to get the the correct
resolution... this might be a possible approach for this.... but I did not found this "feature" like "takeout every second line"
in my gfx-programs....

@OP

Maybe you could share a few sample images and ask people here to convert them to a specific target format AND then share the "recipe" ?

darry wrote on 2024-09-25, 06:58:

@OP

Maybe you could share a few sample images and ask people here to convert them to a specific target format AND then share the "recipe" ?

yes, I will share the original hires monochrome content so that you can have a try.
Resolutions are higher than 720x348, aspect ratio is 1:2,3 or something like that.

This is some contend I would love to convert to: MDA 720x348 2colors and CGA-HIRES 640x200 2colors.
Because the source-material is in a different aspect ratio I would cut it and convert the essential part of the image to the correct pixels and aspect ratio.
Because the patterns are clear and sharp, which indicates that these images are perhaps created on pc and not scanned, the quality can not be obtained when converting to standard MDA or CGA-Hires.

You have to shrink them without using any interpolation.

Pattern is broken...that is the problem.

These are all already dithered images so there is no good way to convert them other than to crop out a section at target resolution.
Any kind of scaling is going to cause moire patterns or other issues within that dithering in original images, it is very much a no-win scenario.

So that is why I thought about taking out every second line and crop the image on the rest.

This will still result in moire patterns against the dither patterns in the original image or changes them entirely. For example with the checkerboard pattern that is prominent in many images, this turns them into vertical lines if every other horizontal line is removed, or horizontal lines if every other vertical line is removed, or into solid color if both every other vertical or horizontal line is removed, and the resulting color depends on pixel offset into the image.

One way would be to de-dither the images, something has to recognize each pattern and assign it some luminance index, scale in that resulting image, and then apply the dither patterns back... but I don't know who is going to come up with the actual algorithms to do it.