Image processing techniques for high speed camera-based free-field optical communication

Optical communication through light-emitting diodes (LEDs) and video cameras is rapidly gaining attention due to the increasing pervassiveness of those devices, and because of its potential data capacity. However, the communication quality is greatly compromised by the low resolution of the imaging sensor which produces blurred images of the LEDs at long distances. On the other hand, the images recorded at high frame rates possess particular features that can be used to improve the reception. This paper suggests image processing techniques to detect and decode the optical signals of an array of LEDs. For the case of blurred images, detection of the LED's is reinforced by a k-means clustering algorithm based on distance measurements derived from the linear correlation among pixel intensities along the time dimension. Experiments with a prototype show that the proposed algorithms can improve the bit error rate (BER) of the decoded signal. Furthermore, partial implementation on a General Purpose Graphics Processing Unit (GPGPU) is also addressed, and processing times are demonstrated.