R&D Computer Graphics
R&D in Silicon Valley. Solved projector keystoning with automated image analysis.
2016
A research and development project in Silicon Valley focused on solving projector keystoning: the trapezoidal distortion that occurs when a projector is not perfectly perpendicular to the projection surface.
The Approach
Working alongside a project manager and a mathematician, I built an Android prototype that displayed calibration patterns and measured the device's tilt in degrees. The prototype projected a white screen in a darkened room, then captured photographs of the projection at various angles.
Image Analysis
Using Python and open-source computer graphics libraries, I analysed each captured image to determine the exact pixel positions of the projected screen's corners. By evaluating how corner positions shifted across different projection angles, we were able to derive a correction algorithm.
The Result
The final algorithm was elegantly simple from the user's perspective: align the projector at 90 degrees to the wall, and the keystoning correction applies automatically. No manual adjustment was needed. The system handled it based on the calibration data we had gathered and the mathematical model we derived.