Showcase: Online Image (Dithering) Converter for E-Paper/LCD

[ningh]

Hi Arduino community!

I've been working on several e-paper display projects recently (monochrome, tri-color, and full-color variants) and encountered a common challenge: converting images for different displays. Existing tools couldn't handle my specific needs like:

• Palette customization for monochrome/limited-color displays
• Real-time comparison of different dithering algorithms
• Direct array output for C/Python embedded systems

After struggling with manual Python + Pillow scripting, I built img2lcd - a pure web-based solution that provides:

• Side-by-side dithering algorithm comparison (Floyd-Steinberg, Atkinson, etc.)
• Custom? palette support (e-papers and RGB565)
• One-click C/Python array generation

Code Sample: (with GxEPD2)

#include <GxEPD2_3C.h>

#include "image.h"  // genreate by img2lcd rgb566_bgr

GxEPD2_3C<GxEPD2_154_Z90c, GxEPD2_154_Z90c::HEIGHT> display(GxEPD2_154_Z90c(/*CS=5*/ 16, /*DC=*/ 15, /*RST=*/ 14, /*BUSY=*/ 13));
void setup() {
  Serial.begin(115200);

  SPI.begin(17, -1, 18);
  display.init(115200);

  display.setRotation(0);
  display.firstPage();
  do {
    display.drawRGBBitmap(0, 0, (uint16_t *)IMAGE, 200, 200);
  } while (display.nextPage());
}

Tested with:
OLED (SSD1306), Good-Display e-papers(1.54"/2.9"/7.5" Mono & BWR), Waveshare 7.3" 7-color e-paper, LVGL (ST7789/RGB565)

I'd Love Your Feedback!

• Have you struggled with image optimization for displays?
• What features would you add (e.g., new dithering modes, palette presets)?
• Found a bug? Let me know!

This tool saved me hours of Python scripting – I hope it helps others too! If you test it, share your experience below or DM me. Happy hacking!

[ftjuh]

Hi ningh,

love that project and it's a shame that you got so little feedback here. Seeing the results of these different dithering algorithms side by side and being able to download a ready to use piece if code is just great.

For my own project, I will implement my own algorithm for ESP32, because I need dynamic conversion, but this has been a great inspiration and helped me choose my preference, which is Atkinson.

After having read up a bit about that algorithm, I wonder how you implemented it for >2 colors, particularly for 3-color B/W/R, as the original only needs to decide between B and W. I tried my own approach (well, frankly, I let chatgpt do the heavy lifting), and got a bit different results than yours, so would love to see your algorithm for that.

Jan

[ningh]

Hi Jan,

Thank you for your kind words! The img2lcd project started as my exploration of dithering algorithms, and it¡¯s still evolving. I¡¯m glad you found the side-by-side comparisons useful.

Regarding multi-color implementations (e.g., B/W/R):

1. Color Matching: Euclidean distance in RGB space determines the closest palette color for each pixel, avoiding simple luminance thresholding.

2. Value Clamping: Post-error accumulation values are clamped to 0-255 instead of redistributing overflows ¨C this might explain our differing results.

If you can share specific test images or code excerpts where our outputs diverge, I¡¯d be happy to analyze them after returning from my trip (currently away from my codebase).

Best,
Ning

[ftjuh]

Hi Ning,

thanks for the explanation. As mentioned, I chose not too dig into these algorithms as deep as you did, and let chatgpt do most of the implementation. As so often, it just needed a more precise prompt and came up with a three-color implementation that did things just fine for my purposes. If you want to have a look, I just put the project on github:

GitHub - ftjuh/ePaper-writer: This ESP32 Arduino sketch lets you write drawn and...

The Atkinson dithering happens in this function, which is a callback from PNG decoding, and thus needs to do the dithering on a line-by-line basis with a three line buffer:

e-paper_writer/e-paper_writer.ino

Anyways, thanks again for your project, it gave me the inspiration to include a dithering option in this project.

Have a safe trip home,
Jan