Detection of Head Motions using a Vision Model

X.W. Gao, S. Batty (UK), L. Podlachikova, D. Shaposhnikov (Russia), and J. Clark (UK)


Motion detection, modelling, PET imaging


Head movement during a brain scan is one of the major reasons causing the blurriness of images. In order to improve the quality of scanned brain images, the head movement parameters should be known and embedded into the image reconstruction algorithms. In this study, a system consisted of two cameras is studied to monitor the head movements under the scanning conditions. The camera is calibrated before the shooting. Images are acquired by the video camera and transferred to a colour monitor that has been calibrated. The range of face skin colours are then obtained based on these images captured under varying lighting conditions to perform face segmentation. The vision model, Foveal System for Face Images (FOSFI) is then applied to find the position of eyes and nose, arriving at the findings of parameters of rotation and translation of head motion. Preliminary results on 2D images with known moving parameters show that movement parameters can be obtained very accurately via the described methods.

Important Links:

Go Back