Motion Detection Using Mean Normalized Temporal Variance

Chan, C W

Lawrence Livermore National Laboratory

关键词: Statistics;    Mirrors;    42;    Image Processing;    Detection;   

DOI  :  10.2172/15004553
RP-ID  :  UCRL-ID-154786
RP-ID  :  W-7405-ENG-48
RP-ID  :  15004553

美国|英语

来源: UNT Digital Library

【 摘 要 】

Scene-Based Wave Front Sensing uses the correlation between successive wavelets to determine the phase aberrations which cause the blurring of digital images. Adaptive Optics technology uses that information to control deformable mirrors to correct for the phase aberrations making the image clearer. The correlation between temporal subimages gives tip-tilt information. If these images do not have identical image content, tip-tilt estimations may be incorrect. Motion detection is necessary to help avoid errors initiated by dynamic subimage content. With a finely limited number of pixels per subaperature, most conventional motion detection algorithms fall apart on our subimages. Despite this fact, motion detection based on the normalized variance of individual pixels proved to be effective.