Yannick Richter's Project Gigapixel Turns an Old Scanner Sensor Into a 200-Megapixel Camera

Computer engineering student and maker Yannick Richter has built a working 200-megapixel camera and is working on extending it to over a gigapixel — by repurposing the charge-coupled device (CCD) from an old Epson scanner.

"While there are a few DIY projects converting CCD scanners into cameras," Richter explains of Project Gigapixel, "most of them use the original scanner hardware and software, essentially packaging a scanner into a new case and reducing the travel range with gears. This usually works, but comes with some limitations mainly in resolution, features, and tricking the original scanner calibrations. This project instead aims to reverse-engineer the usage and protocol of the CCD found in Epson V30(0), V37(0), [and] V200 4800DPI scanners for direct DIY use."

Scanners are designed to capture images of still objects at a far higher resolution than a camera, by slowly scanning them line-by-line — in the case of the sensor used in Project Gigapixel, 12 lines at a time. "The sensor used in these scanners seems to be the 12 line (RGB ×2×2 Main+Sub line) ILX561K CCD [Charge-Coupled Device] with 122,400 total pixels. By using all lines present a square RGB image of over 1.6 Gigapixels with 16b true color could be achievable."

While Richter's current prototype, which connects the sensor to a Raspberry Pi 5 single-board computer with high-speed Non-Volatile Memory Express (NVMe) storage, with the interfacing work offloaded to the programmable input/output (PIO) blocks of the in-house RP1 peripheral chip after proving the concept with an external Raspberry Pi RP2350 microcontroller, doesn't quite reach the target gigapixel-plus, it's already impressively capable — delivering a 200 megapixel image measuring 21,700×10,000 pixels. "There is still slight chromatic aberration," Richter admits of the results so far, "and […] every second line is slightly darker as the sub-lines seem to have a slightly different sensitivity."

The work-in-progress project is documented in full on Richter's Hackaday.io page.