Eye movement correction for 3D OCT volume by using saliency and center bias constraint

The eye movement artifacts occurring during 3D Optical Coherence Tomography (OCT) volume scanning is a problem that easily affect the image analysis and diagnosis. The existing correction methods are influenced by the background noise and strong vessel, which cause the over-correction. To address this problem, we propose an eye movement correction method based on saliency and center bias constraint. Given a 3D OCT volume, our method firstly utilizes the OCT saliency detection to determine the major layer structure for each slice, and assigns a higher weight to the foreground region. Then an image registration with a center bias constraint is employed to estimate the transformation between the neighbor slices, which is employed to wrap the slice based on the first slice of the OCT volume. Our method contains two key insights: (1) applying the OCT saliency detection to extract the layer structure, (2) utilizing the center bias constraint to avoid the distortion caused by vessel matching. Experiments on both synthetic and real datasets show that our method obtain the satisfied results with the saliency and center bias constraint.