Real-time digital image stabilization using motion sensors for search range reduction

To generate sharp and clean picture images, we perform image stabilization. We first acquire the translational and rotational motion vector of the previous frame and the next one in the sequential moving picture. These motion values are used to correct the blurred images. However, in order to determine these values, we are required massive amount of search range and computations. Motivated by these facts, in this paper, we devise a new real-time stabilization technique with taking advantage of both digital image stabilization and motion sensors. The technique uses pitch and yaw values of sensor data for setting the search range, and calculates the translational motion value using SAD (i.e., block matching algorithm) within the search range. Also, roll value is used to compute the rotational motion value. Our experimental results showed that our method successfully performs image stabilization reducing the search range and computational time by 64% compared with the conventional one.