Real Time AOSLO Image Stabilization With Hybrid FPGA-based Hardware/Software Tracking

Abstract

 

Purpose:

 

To test a hierarchical hybrid tracking and image stabilization system combining an FPGA-based beam stabilizer for larger eye movements and an FPGA software stabilization system for fine corrections.

 

Methods:

 

The AOSLO at the Indiana University School of Optometry was used for this study. IU and PSI previously integrated closed-loop retinal tracking for robust AOSLO image stabilization (Ferguson et al, 2010). The PSI-designed "hardware tracker" uses optical sensing of retinal displacement optically and an FPGA-based controller to stabilize to ~5-10µm. Recently, Montana State University and UC Berkeley researchers implemented a novel FPGA-based system (Yang et al, 2010) to map images to preset image templates with µm-level precision. Because the "software tracker" works in a single image frame, it cannot readily compensate large motions. We combined both systems and used the resulting two-stage stabilizer to image four normal volunteers with (a) no tracking, (b)hardware tracking alone, (c)software tracking alone, and (d) the combined trackers, and compared the relative stabilization quality of these methods under a number of test conditions.

 

Results:

 

As the level of tracking increased, the fraction of unregistered frames, and the amount of post-processing required to align, decreased. Figure 1 shows a 1.8 x 1.8 deg image average (both stabilizers running) computed by simply summing 120 frames of video without processing.  

 

Conclusions:

 

Hardware and software stabilization can be directly combined in a two-stage hierarchy. The current system is not a single computer, but demonstrates an unprecedented degree of stabilization for moderate eye movement.

 

Keywords: imaging methods (CT, FA, ICG, MRI, OCT, RTA, SLO, ultrasound) • image processing • photoreceptors