Automatic Keystone Correction in Projector Phones using Computer Vision

Abstract

A Pico projector is an emerging technology which is just now making its way into smartphones. These projectors phones enable consumers to share movies, presentations andvideo-conferences with others by projecting them on nearby surfaces. Projectors whichare not position exactly in front of the projection surface produce images which exhibitthe keystone effect, resulting in skewed and distorted pictures. To compensate for thisthe image is manually corrected by changing properties of the projector. In this thesisa system for automatically correcting a projected image is presented. The system isdesigned to run on a projector phone and uses a camera and accelerometer to calculatehow to correct the image. The algorithm for correcting the image consists of three steps,one of which is to calibrate the system, and this algorithm have been implemented assoftware. A development board with a smart phone processor, pico projector, camera andaccelerometer has been used to run and test the software. The implementation was foundto be slow running, and testing showed that the quality of the correction was not as goodas it could be. This was found to be the result of an inaccurate calibration procedure,and methods for improving it has been provided. It was concluded that a projector phoneusing this system could work consistently, if the calibration was improved.