Electronic image stabilization and video quality assessment in 2-dimension-array

Based on the gray projection principle, the D-subdivision linear interpolation and log-polar transform(LPT) method with a feedback mechanism were presented to estimate the motion vector between reference frame and current frame from an image sequence which included translation, rotation and zoom in 2-dimension-array space. The presented method was used to compensate video image sequence to make the video image sequence output stable. The proposed approach had the translation precision with 1/4 pixel and the rotation precision with 0.2°. Moreover, combining the human visual system with the principle of structural similarity, the gray projection algorithm structural similarity (GPASSIM) was put forward for video quality assessment after image stabilization. The GPASSIM can improve the consistency between the subjective and objective assessments effectively, which has a higher sensitivity than the common used peak signal to noise ratio method.