Non-ideal Conditions For The Simulation Of The Rotational Motion Blurred Image Recovery Method

Image-blurring is caused by the relative motion between the camera and the object during the exposure time. It is also called motion blur. In practice, the motion between the camera and the object is relatively complex. When a camera which works on a moving platform films a target, it often will be accompanied by non-coaxial rotating and swing. This article mainly studies the simulation and restoration algorithms for the rotational motion-blurred image. The main work is given below.1. By analytic geometry and computer graphics, the article studied the mathematical process when a camera works on the non-ideal assumpted conditions (a camera which works on a moving platform does a certain angle of rotation relative to targets, and accompanies by swing on the moving platform), and established the image degradation model. On the base of the research, using rotating angular velocity, sway velocity, exposure time, field of view angle, and image size as parameters, we generated the simulated motion-blurred images under different combinations of the parameters.2. This paper proposed a blur parameter estimation method for rotational motion-blurred images. It divided a rotational motion-blurred image into m×n small images, each image can be approximated to a straight line motion-blurred path. This paper used the way of identifing straight line motion blurred direction and blurred scale to identify the blurred direction and blurred scale for each small image, ultimately formed a rotational motion-blurred field (m×n size) estimate. After that, this paper used the least square method to fit the rotational blurred center and rotational blurred angle. Furthermore, for the estimation method of the rotational motion blur field, this paper did experiments on blurred images from several aspects, and the experimental results were analyzed. Experimental results showed that: when the error between the priori knowledge for the rotational blurred center and the real value was between plus-minus 20 pixels, the error between the least-square estimation of rotating center and the real value was plus-minus 10 pixels.When the error for the priori knowledge for rotational angle was 5°, the least-square estimation of the blurred direction was consistent, and the error compared with the real value was small.3. This paper used a restoration approach based on the polar coordinates transform for the rotational motion blur, did the restoration experiments on the simulated pictures which the rotation center of the blur image was not in the center of the picture. Experiments showed that this algorithm for rotational blurred 5°and 10°images was good, but the bigger therotation blurred angle was, the worse the effect was.